Sales Intelligence Platform - 2027 Vision

Strategic vision for AI-native Sales Intelligence Platform delivering 70% efficiency improvement

Sales Intelligence Platform - 2027 Vision

Prepared by: Jay Chiruvolu, Amplify Sales PM
Date: December 2025
Purpose: Define end-state vision, validate through discovery research, and establish 24-month roadmap for transforming Klaviyo sales operations

Document Overview

This document presents the strategic vision and execution plan for building an AI-native Sales Intelligence Platform that will enable Klaviyo's 600-person sales organization to achieve 70% efficiency improvement (860,000 hours saved over 2026-2027). Based on three weeks of discovery research—including stakeholder interviews across systems teams, sales leadership, and frontline reps, plus hands-on BDR workflow observation—it validates that pipeline generation is the primary constraint (consuming 80% of rep time) and proposes a multi-level solution closely connected to users' most critical needs & pain points.

The vision: By 2027, stateful agents handle administrative work, reps focus on relationships and strategy, and every account has complete context from first touch through renewal. Reps manage 35+ accounts (vs 20 today) with higher quality engagement.

The path: Build intelligence layer on existing systems (Salesforce, Gong, 6sense), leveraging in-progress infrastructure (Customer Master Registry in Snowflake, Data360 interface layer, Project Hermes for BDR foundation). Coordinate with systems teams rather than duplicate, extend work in progress rather than rebuild, deliver value progressively over 24 months.

Table of Contents

Section 1: Ideal Future State (2027)

What the Intelligence Platform delivers. Core capabilities in action.

Section 2: Discovery Findings & Problem Framing

Research synthesis. Pipeline generation is THE bottleneck. Three-question framework. Problems mapped to solution.

Section 3: Required Capabilities

Four layers (Intelligence, Automation, Interface, Learning). Prioritized P0/P1/P2. Dependencies and sequencing.

Section 4: Infrastructure Requirements

Recommended architecture: Snowflake MDM + Data360. Component breakdown. Technology stack.

Section 5: Current State & Critical Gaps

What exists (in-progress work). Critical gaps. What to leverage vs build.

Section 6: Roadmap & Execution Plan

24-month transformation. Year 1: Core BDR capabilities. Year 2: Scale to 600 reps. Monthly breakdown.

Appendix A: Discovery Research - Full Analysis

Complete findings from stakeholder interviews and shadow session.

Section 1: Ideal Future State (2027)

Morning at Klaviyo Sales: What's Different

It's 8:00 AM on a Tuesday in December 2027. Sarah, a Mid-Market BDR, opens her laptop—not to Salesforce, but to the Intelligence Platform. Overnight, agents have been working. They've monitored her 180 workable accounts (out of 500 total in her territory), detected three new intent spikes, researched two companies that got funding, and identified five accounts where champions changed roles. Her morning brief is ready.

The dashboard shows today's intelligent prioritization: 8 accounts flagged as "Act Today" with combined scores above 90 (ICP fit + intent + timing). Each account has a complete intelligence package synthesized from systems she used to check manually: Salesforce, 6sense, LinkedIn, Clearbit, SimilarWeb, BuiltWith, Gong call history, and news. What used to take 30 minutes of tab-switching per account now takes 3 minutes to review because the agent has already aggregated, analyzed, and synthesized everything—including information from call transcripts & other sources she previously would have never seen.

The first priority is Acme Corp. The intelligence brief shows: fashion retailer, $25M revenue, Shopify Plus (perfect fit), Mailchimp contract expires in 4 months, new CMO hired 2 months ago (likely reviewing stack), LinkedIn activity shows frustration with attribution, 6sense detected three searches for "marketing attribution" this week, pricing page viewed three times. The agent has identified the CMO (Sarah Chen), verified her mobile number through Clay waterfall enrichment, and drafted a personalized email referencing her LinkedIn post about attribution challenges—with the same completed for 10 other potential contacts. Sarah reads the brief, makes a small edit to the email (adds a case study reference), and approves. The agent sends it. Three minutes total.

By 5pm, Sarah has worked through 15 accounts and booked 6 meetings. In the old world, she'd have booked 1-2 meetings after researching 8-10 accounts. The difference is that research happens continuously by agents who never sleep, synthesize signals humans can't manually connect, and present her with ready-to-execute plans rather than raw data to interpret.

The Intelligence Platform: Core Capabilities

1. Intelligent Daily Prioritization

Every morning, the system answers "What should I do today?" The priority queue is ranked by machine learning combining ICP fit, intent signals, and timing factors. It refreshes continuously—if an account shows intent spike at 2 PM, it moves to the top immediately.

Role-specific priorities:

• BDRs: Accounts by conversion probability (fit + intent + timing + workability)
• AEs: Deals by health + urgency + value (what needs attention now)
• CG AEs: Customers by expansion readiness + churn risk
• Managers: Team priorities with root cause analysis and coaching recommendations

Cross-system synthesis example: BeverageCo jumped to top priority at 10 AM when the agent detected someone forwarded your teammate's email to a VP, the VP opened it three times, and LinkedIn confirmed the VP as the decision-maker. Action: "Call within 2 hours while engaged."

2. Complete Account Context (3-Minute Research)

Click any account and get instant comprehensive intelligence:

Company Profile: Industry, revenue, employees, growth trajectory, funding status (6 sources aggregated)

Tech Stack: E-commerce platform, current providers, contract expirations, pain points (BuiltWith + integrations)

Intent & Timing: Intent score with specific search terms, website behavior, LinkedIn activity, budget cycle timing, contract renewal windows (6sense + website + LinkedIn + news)

Buying Committee: Decision-maker hierarchy, influence patterns, past engagement, relationship strength scores, recommended entry point (LinkedIn + org chart + Gong history)

Past Engagement: Every call with transcript links and key moments, email exchanges with sentiment, demos with feedback, objections and how addressed (complete history with citations)

Similar Customer Playbook: Comparable wins with key factors, common objections, typical deal structure, success metrics to reference (pattern matching)

Why now synthesis example: Global Retail Corp shows moderate intent (78/100) but high value ($85-120K). The agent connects: Q1 budget planning season, fragmented attribution across acquired brands (LinkedIn CEO post), expensive incumbent contract, GDPR hiring signal. Recommended strategy: Lead with multi-brand attribution demo, reference similar customer case study, target VP Marketing.

3. AI-Drafted Content (Review, Not Create)

Agents draft everything. You review and approve:

Outreach: Emails and call talk tracks personalized to account intelligence, learned from your edits over time, includes relevant case studies

Meeting materials: Prep briefs 30 min before calls, post-call follow-ups with ROI calculators, demo scripts customized to their use case

Proposals: Pricing based on usage profile and similar customers, terms aligned to preferences, ROI projections using their stated metrics

Time transformation: What used to take 15-20 minutes of drafting per item now takes 2 minutes of review. You focus on strategy—does this positioning resonate? Should we multi-thread to CFO?—while agents handle tactical execution.

4. Automatic Administrative Work

You never touch Salesforce directly. Agents handle all updates:

After every call: MEDDPICC fields extracted with confidence scores and citations, stage progression updated, next steps logged, follow-up tasks created

After every email: Response tracked, sentiment analyzed, engagement level updated

After every meeting: Attendance tracked, no-shows flagged, rescheduling coordinated automatically

Activity logging: Perfect history without manual entry. CSM receiving handoff sees complete story without AE repeating everything.

The result: Your CRM time goes from 60 minutes per day to zero. You review what agents updated (5 minutes spot-check) but don't do any data entry.

5. Real-Time Intelligence During Calls

You're on discovery call with RetailCo. At 8:32, Lisa mentions they're evaluating Braze. Instantly, the agent:

• Surfaces Braze battlecard with positioning angles
• Shows you what to say: "Braze is powerful but complex—most mid-market teams find setup takes 3-4 months vs our 30 days"
• Notes this for MEDDPICC Competition field

At 12:05, someone new joins the call. Instantly, the agent:

• Identifies: John Smith, CFO
• Flags: Not on original invite (escalation signal—they're taking this seriously)
• Provides: CFO profile (cost-conscious, mentioned ROI three times in past emails)
• Suggests: Adjust approach to emphasize cost savings

Post-call, you do nothing. The agent already extracted all MEDDPICC fields, updated Salesforce, drafted follow-up email with ROI calculator, scheduled demo, and generated next best actions.

Your time: 45-minute call + 5-minute review = 50 minutes total (vs 45-minute call + 60 minutes of admin today).

6. Proactive Risk and Opportunity Detection

Agents monitor continuously, catching signals before they become problems:

Champion changes: Detect engagement drops (Jamie hasn't opened last 3 emails, canceled meetings) + track role changes (promoted on LinkedIn) → Identify replacement (Sarah Kim, new Marketing Director), draft transition emails, recommend immediate outreach. Without agent: Deal stalls when champion goes MIA, discovered too late.

Expansion timing: Track stated goals ("expand to SMS eventually" from 6 months ago) + monitor readiness (list growth, budget cycle) + similar customer patterns (85% attach at 6-9 months) → Draft expansion proposal. Without agent: Opportunity sits in "eventually" until competitor suggests it.

Churn risk: Health signals (send volume down 30%, support tickets up 3x) + relationship signals (champion not responding) + undelivered commitments → Coordinate intervention with specialists. Without agent: Surfaces 30 days before renewal when recovery rate drops to 20%.

The impact: Nothing falls through cracks. Agents catch and respond to signals humans miss or discover too late.

7. Persistent Memory & Context Retention

Territory intelligence survives handoffs: Jenna (SMB BDR) moves into an AE role. Her territory knowledge—warm accounts, follow-up timing, champion relationships, pain points—currently lives in her head. With the Intelligence Platform, that knowledge persists. The new BDR sees every conversation with key moments highlighted, follow-up requests auto-surfaced ("RetailCo said check back December 2025" → appears in December queue), and relationship strength built over months.

Insights surfaced, not buried: Today, important information lives in 500+ Salesforce activities per account. With persistent memory, agents continuously extract insights. When Bill (ENTR AE) opens a churned account to re-engage, he sees the actual context: deliverability issues for 3 months, support couldn't resolve, attribution setup never completed, champion left frustrated. Recommended approach: Lead with "we've improved deliverability," acknowledge past issues.

From shadow: "75% of my book is churned customers, but support ticket history is completely invisible. I have no idea why they left."

Knowledge capture from humans: Territory insights in rep brains and spreadsheets get captured automatically—tier 1 target lists become Intelligence Platform priorities, "this contact is the real decision-maker" becomes buying committee intelligence, seasonal timing patterns become optimization rules.

The transformation: Context becomes organizational asset, not personal knowledge. Both humans and AI agents benefit from accumulated intelligence that doesn't disappear when someone changes roles.

Manager Intelligence: Strategic, Not Administrative

Emma (Manager) reviews team dashboard:

Pipeline health with root cause: Mark at 2.9x coverage (below 3.5x target). Agent identified root cause: Connect rate 22% vs team 35%. Diagnosed WHY: Mobile number coverage 60% (vs Sarah 85%, Lisa 82%). Recommended fix: Enable Clay waterfall for Mark's accounts. Auto-generated 1:1 agenda ready.

Performance pattern detection: Lisa crushing it at 40% self-sourced pipeline (vs team 20%). Agent analyzed her approach: Uses auto-sourcing alerts, acts within 24 hours. Recommendation: Replicate Lisa's playbook across team. Action: Agent drafted team meeting agenda featuring Lisa.

Deal risk flagging: GlobalCorp at risk—champion lost (detected via cross-system signals). Agent provided Mark's multi-threading plan. Emma reviews, approves. Follow-up queued.

What Emma doesn't do: Manually review 80 team deals, dig through Salesforce for context, guess at coaching needs. What Emma does: Strategic leadership, data-informed coaching, pattern replication. 30-45 minute review vs 2-3 hour admin burden today.

Why This Is Achievable in 2 Years

The capabilities described aren't science fiction, just smart engineering:

Data synthesis across systems is technically straightforward (API integrations + data normalization + LLM summarization). Cross-system signal detection is pattern matching (if LinkedIn shows promotion AND email engagement drops, flag champion risk). Content drafting uses LLMs that already exist. Automatic CRM updates are API calls based on extracted data.

What makes it realistic:

• Building on existing tools (Salesforce, Gong, 6sense stay as data sources)
• Proven vendor solutions for hard parts (state compression, agent orchestration)
• Incremental delivery (pipeline intelligence Month 4, full platform Month 24)
• Similar systems exist today (Actively AI running millions of agents in production)

What makes it transformational:

• Unified intelligence layer where everything compounds (not isolated features)
• Every capability shares the same Customer 360 foundation
• Agents learn from every interaction, improving continuously
• The whole is greater than sum of parts

Section 2: Discovery Findings & Problem Framing

See Appendix A for detailed overview of insights from research

What We Learned (3-Week Discovery)

Interviewed 11 stakeholders across systems teams, sales leadership, enablement, and frontline reps. Shadowed BDRs through complete daily workflows. Combined top-down strategy perspective with bottom-up operational reality.

The core insight: Pipeline generation consumes 80% of rep time and is THE constraint on efficiency—not closing skills, not deal management, but finding and engaging the right prospects.

From interviews: "Pipeline is the biggest issue. It's actually kind of impossible to hit quota for most reps because there's not enough demand."

Why this matters: Traditional coverage requires 3-4x pipeline to quota. With relatively fixed close rates and limited revenue calls per day (2-3 hours max), need more opportunities to work, not just better closing.

The Pipeline Generation Workflow (Three Core Activities)

Shadow session and BDR director interviews revealed consistent framework. Every BDR/AE asks three questions daily:

1. What accounts should I go after today? (Account prioritization)
2. Who are the right people at those accounts? (Contact identification)
3. What should I say to them? (Outreach personalization)

These three activities consume the 80%. Understanding problems in each reveals what to build.

The Problems

Activity 1: Account Prioritization

No systematic account scoring. Morning routine takes 45-60 minutes checking 6sense, Salesforce, and deciding priorities manually. From ~200 workable accounts, picking 10-20 requires experience-based judgment synthesizing intent + timing + firmographics across disconnected tools.

Territory workability crisis: 70% of assigned territory is unusable (customers not marked, invalid URLs, duplicates, B2B non-fits, parent-child confusion). BDRs manually categorize every account.

Timing intelligence buried. Contract renewal data exists but not surfaced. From shadow: "Find an account that says their contract ends December 2025 and I'm like, how did I not know that? Why was that not on some report for me? That was pure luck."

Static prioritization. Morning decision creates fixed list for the day—no reprioritization if signals change. Intent spike at 2 PM doesn't surface. Champion response doesn't trigger next steps automatically.

Current tools: 6sense, Salesforce, Similar Web, BuiltWith, Clearbit, ZoomInfo, Google. Manual synthesis required.

Activity 2: Contact Identification

Broad search easy, prioritization hard. LinkedIn Boolean filter produces 30-40 relevant people in 5 minutes. Ranking them takes 10-15 minutes using experience-based pattern recognition. Early-career BDRs lack the 6+ months of mental models needed to assess decision authority and champion potential.

Deep account penetration not systematic. BDRs contact 2-5 people per account average. Ideal state: 50 prospects per account, cascading automatically through organization. Southwest example: 239 identified prospects, only 2-3 contacted. 100x untapped potential.

Past engagement context buried. To check if contact was previously reached: navigate to Account → Activity tab → scroll through all activities → search for name. Time-consuming, easy to miss.

Contact quality varies. Enterprise contacts accurate, SMB has issues (wrong numbers, bounced emails, stale LinkedIn data showing previous companies).

Current tools: LinkedIn Sales Navigator, Salesforce, Gong Engage, ZoomInfo. Knowledge scattered.

Activity 3: Outreach Personalization

AI email GPT requires iterations. Current process: 20-25 min per email (vs 30-40 without AI). Team-wide agent produces first draft but requires 3-5 refinement cycles due to generic voice, formulaic structure, weak calls-to-action, and insufficient account context. Quality gap: Current = "better than I could write" but needed = "perfect first draft, minimal edits."

Research scattered. LinkedIn profiles, Salesforce closed-loss notes (requires navigation), manual account maps in Google Sheets. Rep aggregates from 5+ sources over 10-15 minutes. No automated synthesis.

Knowledge not captured. From shadow: "My notes app is so disorganized—million random things I think of. Notes are in my head or scattered. If I quit, they're lost forever." Account intelligence lost with every departure.

Current tools: LinkedIn, Salesforce, Google Sheets, Notes app, ChatGPT, Gong Engage. Manual aggregation.

Cross-Cutting Issues

Tool Sprawl & Integration Gaps

15+ distinct tools used daily across the three activities. Impact beyond time investment:

Cognitive switching costs: Different interfaces, search patterns, authentication sequences.

Information silos: 6sense intent not connected to Salesforce closed-loss context. Rep must correlate signals mentally across disconnected systems.

Navigation tax: Knowing exact path to find information (Activity tab for calls, Opportunities for closed-loss). Hundreds of possible Salesforce reports to review. New (and experienced) reps get lost.

Follow-up chaos: Zoe (SMB BDR) works across Gong (tasks) + Gmail (thread view) + Calendar (reminders) to manage follow-ups. "I would do anything for a tab in Gong showing accounts by days since last touch: 1 day, 2 days, 3 days, 4 days, 5+ days. Currently memory-dependent—I remember because I have a good memory."

Context loss: LinkedIn research not captured in Salesforce, insights disappear.

From shadow: "We pride ourselves on being consolidated platform, but to work here we use 20 platforms. Should have platform like Klaviyo for our own process."

Data Quality & Trust Crisis

Current state: "We are behind where companies were five years ago. Our CRM is rampant with misinformation. To find a prince, you have to kiss 7 million frogs daily."

Trust breakdown pattern:

1. Tools display inaccurate or outdated data
2. Reps try once, encounter errors
3. Permanent rejection: "Used it once, was wrong, never touched again"
4. Cannot recover even after improvements

Visibility gaps: Maddie (ENTR AE) has 75% churned customers in her book, but support ticket history completely invisible. "Did they have the worst support experience ever? I didn't see that. I have no idea why they left." Critical context missing when re-engaging.

Specific issues: Field mapping errors (Lead → Contact data loss), country field problems (breaking enrichment), enrichment clutter (5 providers creating redundant fields—"Which revenue is correct?"), hierarchy staleness, deduplication backlog, 500+ fields accumulated over 10 years.

What reps trust vs ignore:

• Trusted: Closed-loss notes ("gold standard"), activity logging, account hierarchy
• Untrusted: Enrichment fields, account ownership, contact routing—anything not personally verified

The adoption requirement: 95% accurate = 5% wrong = rejected by burned team. Need 98-99% accuracy for adoption. From interviews: "Only way forward: do really meaningful stuff that works from day one."

Organizational Dysfunction

Sales/Systems relationship strain: "There is such an unhealthy feedback loop between sales and systems. It's the worst it's ever been."

Process failure: Sales leader identifies need → sales ops creates ticket → ticket enters backlog → systems team doesn't see request for 8 months (described as "graveyard"). Visibility breakdown: sales unaware of systems roadmap, systems unaware of sales pain points. Communication structure loses information in translation.

Governance Vacuum: "Company's AI push good, but led to not as much PM support and governance. It's lot of people building—let's build first, figure out later if it's coherent." Six teams building simultaneously (Hermes, CMR, Data360, Enrichment, Lead scoring, Amplify) with no coordination mechanism. Risks: duplication, incompatibility, wasted resources.

Tool acquisition dysfunction: Operations buys new tools, systems finds out after. "We're thinking 'we already have 7 tools that do that. Wish you brought us along for conversation.'"

Engineering filling product gap: "Without strong product team, work lands on engineers. We're not product people. Work feels random—tackling initiatives other teams lack capacity for. Not sustainable."

Reporting & Analytics Vacuum

700 individual reports per person. No core dashboards. Salesforce dashboards break when someone overwrites. Gong reporting described as "trash."

Manager challenges: Every manager builds own in Excel or custom Salesforce reports. Each BDR/AE maintains hundreds of personal reports. Team dashboards fail: created for group → someone overwrites → broken for everyone.

Can't answer basic questions: "What was bounce rate in January vs now?" No health metrics over time, no trend analysis, no root cause identification.

Impact: Can't coach with data (guessing vs root cause), can't identify patterns (what messaging works), can't replicate top performers (one rep self-sources 40% pipeline vs team 20%—how?), can't make informed strategy decisions (intuition vs metrics).

FY26 operational challenge: BDR:AE ratio changing to 1:4. Need attribution visibility across four AE relationships. Need coverage tracking: out of workable accounts, how many actively worked?

Product Data Access Gap

"If we can get APIs to product data, that'll be huge—it's a huge gap across the board."

Cross-functional need: Sales (cross-sell signals, health scores), Marketing (in-app behavior), CS (expansion opportunities, churn risk). Product usage data exists in Snowflake but no easy sales access.

Impact on intelligence: Cannot predict expansion without usage signals (hitting limits? using advanced features?), cannot detect churn risk without engagement trends (declining sends?), cannot identify whitespace without adoption data (missing products they should have).

Validation: Shadow Session Confirms Solution

When asked for ideal state, the BDR immediately described our Section 1 vision:

"In an ideal world, every day I come in, AI gives me 10 accounts prioritized, which people have best chance responding, what to message them."

Her three questions map directly to the solution:

• What accounts? → Account prioritization intelligence
• What people? → Contact prioritization intelligence
• What to say? → AI-drafted messaging

She's already envisioning what we're building. This is the strongest possible validation.

Section 3: Required Capabilities

The problems in Section 2—no systematic scoring, tool sprawl across 15+ systems, data quality crisis, organizational dysfunction—reveal the capabilities needed to transform sales operations. Pipeline generation is our first priority, making BDR capabilities the Year 1 beachhead. But the same Customer 360 foundation that enables BDR intelligence also supports AE deal progression, Manager team intelligence, and CG AE post-sale capabilities—full lifecycle coverage by Month 24.

This section details capabilities by persona (BDR → AE → Manager → CG AE), then shows platform capabilities that enable all personas, and concludes with prioritization (P0/P1/P2) and dependencies.

BDR Capabilities: Pipeline Generation (Year 1 Beachhead)

These capabilities answer the three questions consuming 80% of BDR time:

Question 1: What Accounts Should I Go After Today?

Account Prioritization: ML-ranked daily list combining ICP fit, intent signals, and timing factors. Refreshes continuously—intent spike at 2 PM moves account to top immediately. Replaces 45-60 minute morning routine with 5-minute review of intelligent priorities.

ICP Fit Scoring: 0-100 score based on firmographics (revenue, employees, industry), technographics (e-commerce platform, tech stack sophistication), and behavioral signals (website engagement, content downloads). Enables systematic multi-factor scoring vs manual experience-based judgment.

Intent Signal Aggregation: Synthesizes 6sense, website behavior, LinkedIn activity, news, and funding into unified intent score with specific triggers. Surfaces "why now" by connecting multiple weak signals into strong composite (contract timing + new CMO + attribution frustration = ready to buy).

Timing Trigger Detection: Automatically surfaces contract expirations, budget cycles, stakeholder changes, funding events. Solves discovery gap where critical timing data exists but isn't surfaced—from shadow: "How did I not know their contract ends December 2025? That was pure luck."

Territory Validation: Instant answer to "Can I work this account?" Checks customer status, deduplication, parent-child relationships. Automates the manual categorization BDRs currently do to identify the 70% unusable territory (customers not marked, duplicates, B2B non-fits).

Inbound Lead Processing: Automatic qualification and routing of inbound leads based on fit and intent. Eliminates manual triage time.

Question 2: Who Are the Right People at Those Accounts?

Contact Prioritization: Within-account ranking of 30-40 LinkedIn contacts by decision-making authority, past engagement, and champion potential. Reduces 10-15 min manual analysis to automated ranking. Uses pattern recognition early-career BDRs lack.

Buying Committee Analysis: Decision-maker hierarchy, influence patterns, multi-threading strategy. Enables systematic deep account penetration—239 prospects per account vs current 2-3 contacted (Southwest example from shadow).

Contact Intelligence: Auto-populate contacts from LinkedIn, enrich with verified emails/mobiles, maintain currency as people change roles. Surfaces past engagement history without navigating to Activity tab and scrolling through 500+ activities.

Question 3: What Should I Say to Them?

Account Research Automation: Pre-generated briefs synthesizing company profile, tech stack, pain points, timing triggers, competitive context, similar customer playbooks. Aggregates what currently takes 10-15 minutes across 5+ sources. Delivers 3-minute research vs 30-minute manual.

AI-Drafted Outreach: Personalized emails and call talk tracks that learn individual rep voice from edits over time. Addresses current quality gap—team agent produces "better than I could write" but requires 3-5 iterations. Target: strong first draft with minimal edits (20-25 min → 5 min).

Automated Knowledge Capture: Territory insights, contact notes, messaging strategies captured automatically as workflow happens—not in rep's head or scattered Notes app. From shadow: "If I quit, lost forever." Enables territory intelligence to survive handoffs.

AE Capabilities: Deal Progression

MEDDPICC Auto-Extraction: Extracts all 8 MEDDPICC fields from Gong discovery/demo call transcripts with confidence scores and citations. Auto-updates Salesforce. Eliminates significant CRM update time.

Deal Health Monitoring: Real-time assessment from engagement trends (email opens, calendar patterns, call sentiment). 7-14 day early warning before deals stall vs reactive discovery at forecast calls.

Win Probability: Continuous prediction (±15 points accuracy) pattern-matching historical deals. Enables accurate forecasting and risk detection. Current: rep gut feel often 30+ points off.

Automatic CRM Updates: Stage progression, next steps, activity logging after every interaction—agent writes to Salesforce. CRM time 60 min/day → 0.

Meeting Coordination: Scheduling, invites, optimal timing, attendance tracking, no-show flagging, automatic rescheduling.

Follow-Up Automation: Post-call materials (ROI calculators, case studies) sent automatically with context-aware content. Never miss a follow-up.

Proposal Generation: Pricing based on usage profile and similar customers, terms aligned to preferences, ROI projections using their metrics. 2 hours → 15 min review.

Manager Capabilities: Team Intelligence

Manager Dashboard: Team performance with root cause analysis (Mark's 22% connect rate vs team 35%—WHY? Mobile coverage 60%). Coaching recommendations data-driven and specific per rep. Replaces poor Salesforce/Gong reports and 2-3 hour manual review with 30-45 minute strategic session.

Performance Analytics: Identify what top performers do differently (one rep 40% self-sourced pipeline vs team 20%—how?). Extract patterns, replicate playbooks across team. Addresses current inability to coach with data.

Coaching Recommendations: Skill gap identification, training needs, best practice suggestions based on call analysis and outcome patterns.

CG AE Capabilities: Post-Sale Intelligence

Expansion Propensity: Readiness signals from product usage patterns + similar customer timelines. Flags opportunities 2-4 weeks early. Current: CSM notices usage increase, maybe mentions.

Churn Risk Prediction: 60-90 day early warning with root cause (declining usage + support tickets + engagement drop). Intervention at 60% recovery rate vs 20% at 30 days. Surfaces invisible context (support ticket history) that Maddie can't see today.

Platform Capabilities: Foundation for All Personas

Customer 360 Platform: Unified data layer with canonical IDs (via CMR), linking all customer data across Salesforce, Gong, 6sense, Zendesk, product usage. Solves information silos—enables cross-system synthesis that humans can't do manually. Foundation for all intelligence capabilities.

Data Quality & Monitoring: Continuous validation, accuracy tracking, deduplication. Achieves 98-99% accuracy requirement (95% = rejected by burned team). Single source of truth per data point.

Command Centers: Role-specific interfaces (BDR, AE, Manager, CG AE) consolidating 15+ systems into one daily workspace. Morning brief, complete account context, approval interface. From shadow: "Should have platform like Klaviyo for our own process."

Notifications & Alerts: Critical signals (intent spikes, champion changes, competitor mentions, deal stalls) delivered in Slack where reps work. Daily digests (morning priorities, evening summary).

AI Assistant: Natural language queries for ad-hoc analysis ("Which accounts should I prioritize this week?"). Context-aware, knows what you're working on. Addresses reporting vacuum—conversational queries vs 700 individual reports.

Continuous Learning: Email response pattern learning, call timing optimization (Keegan discovered West Coast 5-8 PM = 20% connect vs <5% during day), deal progression playbooks, ICP model evolution. Platform gets smarter with every interaction—network effects create competitive moat.

Capability Prioritization & Dependencies

P0 - Foundation (Months 1-6)

BDR pipeline generation capabilities + Customer 360 Platform:

• Customer 360 Platform, Data Quality & Monitoring
• Account Prioritization, ICP Fit Scoring, Intent Signal Aggregation, Timing Triggers, Territory Validation, Inbound Lead Processing
• Contact Prioritization, Buying Committee Analysis, Contact Intelligence
• Account Research Automation, AI-Drafted Outreach, Automated Knowledge Capture

Why first: Pipeline generation is strongest PPR lever. Build BDR beachhead with high accuracy (98-99%) to rebuild trust with burned team. Foundation enables all other personas.

P1 - Transformation (Months 4-15)

AE deal progression + Manager intelligence + automation:

• MEDDPICC Auto-Extraction, Deal Health Monitoring, Win Probability
• Automatic CRM Updates, Activity Logging, Meeting Coordination, Follow-Up Automation
• Command Centers (all personas), Notifications
• Manager Dashboard, Performance Analytics, Coaching Recommendations
• Proposal Generation, Workflow Orchestration

Can start early: Email drafting (Month 4) only needs account intelligence. Basic Command Center (Month 8-9) shows intelligence as soon as it exists. MEDDPICC extraction (Month 5-6) leverages Gong integration work.

P2 - Polish (Months 13-24)

Post-sale intelligence + continuous improvement:

• Expansion Propensity, Churn Risk Prediction (require product data integration Month 13+)
• Continuous Learning (requires volume of interactions for pattern detection)
• AI Assistant (nice-to-have, not blocking transformation)

Principle: Build BDR foundation (Months 1-6), expand to AEs and Managers (Months 7-15), add post-sale and continuous improvement (Months 13-24). Phased value delivery while building toward full lifecycle intelligence.

Section 4: Infrastructure Requirements

The capabilities in Section 3—from BDR pipeline generation to Manager team intelligence—require infrastructure organized into five components: Data Platform (unified customer data), Intelligence & Agent Infrastructure (AI capabilities + system to run them), Integration & Automation (connect systems + execute decisions), Interface (human control), and Observability (measure and improve). Each component enables specific capabilities from Section 3.

Component 1: Data Platform (Customer 360 Foundation)

Purpose

Unified data layer linking all customer information across Salesforce, Gong, enrichment data, Zendesk, product usage, and support—from first touch through renewal. Enables cross-system synthesis that humans can't do manually (Section 2 problem: 15+ systems, information silos, manual correlation required).

Requirements

Identity & Hierarchy:

• Canonical customer IDs (one legal entity = one account across all systems)
• Parent-child hierarchy (enterprise accounts with subsidiaries)
• Identity resolution (same person/company across different identifiers)
• Solves: Parent-child confusion, deduplication chaos, 70% unusable territory

Data Storage & Access:

• Complete account profiles (firmographics, technographics, contacts, buying committee)
• Interaction history (calls, emails, meetings, support tickets, product usage)
• Intelligence data (intent signals, ICP scores, win probability, health metrics)
• Query performance: <200ms for agent queries at scale (500K-1.5M accounts)
• Storage: 5+ year history per account without degradation

State Compression & Retrieval:

• LLM context problem: Can't process 5 years of transcripts per query
• Compressed "base state" (core facts + insights) + raw data on demand
• Enables: 3-minute research vs 30-minute multi-system navigation

Two Viable Options

Option A: Data360 (Path of Least Resistance)

What it provides:

• Clean interface layer for business users across company (sales, support, success, marketing)—not just a data warehouse, but business-user-accessible platform
• Robust APIs with set schemas/metadata for building AI applications—extract single consolidated data object from multiple systems cleanly
• Zero Copy from Snowflake (all data accessible without ETL)
• Vector search built-in (LLM context retrieval)
• Native integrations (Salesforce, Snowflake, Zendesk) + Agentforce enabled

Advantages:

• Faster time-to-value (leverage existing vendor capabilities)
• Less build effort (don't need to create interface layer ourselves)
• Easier to build AI on top (clean APIs, established patterns)
• Cross-functional benefit (not just sales—support and success get data access)

Concerns:

• Cost model unclear (pay-per-usage at 500K-1.5M accounts with high query volume)
• Less control over data model and features
• Vendor dependency

Timeline: Apparently committed as platform is Agentforce dependency—could configure for Amplify needs Q1 2026

Option B: In-House Snowflake Customer 360

What it provides:

• Full Customer 360 capabilities built in-house on Snowflake
• Complete control over data model, features, and evolution
• CMR (canonical IDs, hierarchy, identity resolution) already in progress

Advantages:

• Lower cost (no vendor licensing, just Snowflake compute)
• More control over features and data model
• No vendor dependency or lock-in

Concerns:

• Onus on us to build interface layer for business user access (Data360 provides this, we'd need to build it)
• Slower time-to-value (build vs buy)
• Need to create the clean APIs for AI application development ourselves
• Maintenance burden (data model evolution, schema management)

Timeline: GrowthArc team building Q1-Q2 2026

Both options support Intelligence Platform needs. Key differences: build effort vs control, speed vs cost, vendor solution vs in-house. Decision depends on GTS Data capacity, timeline urgency, and total cost comparison.

CMR provides canonical IDs for either option—identity layer works regardless of interface choice.

Validation Needed (Q1 2026)

For Data360:

• Query performance at 500K-1.5M accounts with high agent query volume
• Total cost model (not just base licensing—usage, API calls, storage at our scale)
• Can we build custom intelligence on APIs effectively?
• Timeline alignment with CMR delivery

For In-House Snowflake:

• GTS Data capacity and timeline (can they build interface layer by Q2 2026?)
• What APIs/interface would they provide for AI applications?
• Query performance at scale with in-house implementation
• Total build effort required from Amplify team

Decision timing: Month 2-3 after validation testing

Component 2: Intelligence & Agent Infrastructure

Purpose

AI capabilities that enable Section 3 outcomes (account prioritization, contact intelligence, MEDDPICC extraction, win probability, etc.) + infrastructure to host and orchestrate agents at scale (500K-1.5M accounts).

Intelligence Capabilities (We Build - Our Differentiation)

Scoring & Prediction Models:

• ICP fit scoring (Klaviyo-specific conversion patterns)
• Win probability (trained on our deal outcomes)
• Champion strength (our relationship indicators)
• Expansion propensity, churn risk (our usage → outcome patterns)

Research & Synthesis:

• Account research automation (synthesize 6+ sources)
• Competitive intelligence (account-specific positioning)
• Buying committee analysis (influence mapping)
• Next best action recommendations (our workflow logic)

Why we build: Intelligence improves with Klaviyo-proprietary data. Vendors have generic scoring; we need our patterns. This is our competitive moat.

Agent Infrastructure (Buy or Build)

The challenge: Need to host 500K-1.5M agents (one per account/lead), each maintaining compressed state, continuously updating, coordinating workflows. Building this ourselves = months of infrastructure work before delivering any capabilities.

Option A: Actively AI (Likely Path - POC in Progress)

What it provides:

• Agent hosting at scale: 1M agents running continuously (proven with millions in production)
• State management: Compressed account memory per agent, append-only facts with citations, API access to query state
• Agent Inbox (push): Daily delivery of top prioritized accounts with complete research brief, contact prioritization, pre-written outreach—one-click add to Gong Engage with full sequence
• Assistant (pull): On-demand queries via Slack or Actively UI for any account ("Help me break into RetailCo")
• API for custom development: Query account state, scores, research—enables building custom interfaces and more advanced intelligent applications
• Base integrations: Salesforce, Gong, Gong Engage, Snowflake, Slack (native connectors)

How we use it:

• Actively provides: Infrastructure (agent hosting, State, delivery mechanism, APIs)
• We build: Klaviyo-specific intelligence on their SDK (ICP scoring, competitive positioning, "why now" synthesis)
• We configure: Prioritization logic, scoring weights, distribution, & interface

Advantages:

• Accelerates from "can't host an agent" to "1M agents at scale" (2-3 months vs 6-12 months custom)
• Full delivery mechanism (Agent Inbox wraps research + contacts + outreach for high adoption)
• APIs enable extensibility (not locked into their UX—can build custom Command Centers)
• Proven at scale with similar sales orgs

Concerns:

• Cost at 500K-1.5M scale (low seven figures annually—validate ROI justifies expense)
• Nightly batch refresh (not real-time continuous during POC, though future triggers possible)
• Customization depth (can we build sufficiently differentiated Klaviyo intelligence on SDK?)

POC validation (Months 1-2):

• Track 1: BDR value (70%+ acceptance rate, conversion ≥ manual selection, productivity lift)
• Track 2: Extensibility (API enables custom interface development, foundation for future use cases)

Option B: Custom Agent Infrastructure (Fallback)

What we'd build:

• LangGraph for orchestration
• PostgreSQL + pgvector for state storage
• AWS infrastructure for agent hosting
• Custom event system for coordination

Advantages:

• Full control over architecture
• No vendor dependency
• Potentially lower long-term cost

Concerns:

• 6-12 month build before delivering any capabilities
• 3-5 infrastructure engineers required (vs 2-3 with Actively)
• Ongoing maintenance burden

Decision: Validate Actively in Months 1-2. If POC successful, use as accelerator. If constraining or cost-prohibitive, build custom.

Component 3: Integration & Automation

Integration Layer (Connect the Senses)

Purpose: Pipe data from source systems into Customer 360 Platform

Most integrations already exist:

• Source systems → Snowflake (Salesforce, Zendesk, product usage, NetSuite via nightly batch)
• Gong → Data Cloud in progress (Systems team)
• Enrichment Service launching Jan 15 (Systems team)

What Amplify must build:

• LinkedIn monitoring (champion job changes, company posts, activity signals)
• Email/calendar tracking (engagement metrics, response patterns, meeting attendance)
• News/funding APIs (external timing signals)

Automation Layer (The Muscles)

Purpose: Execute Brain's intelligent decisions automatically

Capabilities:

• CRM automation (MEDDPICC updates, stage progression, activity logging → Salesforce)
• Communication automation (email sending, meeting scheduling via Gong Engage, calendar APIs)
• Workflow orchestration (SA coordination, demo prep, proposal generation)

Build approach: APIs + Workato for workflow logic, approval gates for customer-facing actions

Component 4: Interface & Experience

Purpose

How humans consume intelligence and control agents—consolidate 15+ systems into role-specific workspaces.

Requirements:

Command Centers (React web app):

• Persona-specific: BDR, AE, Manager, CG AE views
• Daily priorities, complete account context, approval interface
• Fast (<2 sec load), responsive, mobile-friendly

AI Assistant:

• Natural language queries ("Which accounts should I prioritize?")
• Addresses reporting vacuum (conversational queries vs 700 individual reports)

Notifications:

• Critical alerts (Slack where reps work)
• Daily digests (morning priorities, evening summary)

API for extensibility:

• Enable other teams to build on intelligence
• Executive dashboards, BI reports

Component 5: Observability & Continuous Improvement

Purpose

Measure accuracy, track performance, improve models—critical given 98-99% accuracy requirement for burned team.

Requirements:

Evaluation Infrastructure:

• Eval datasets (ground truth for ICP scores, win probability predictions)
• Accuracy tracking (predicted vs actual outcomes over time)
• Human feedback collection (rep corrections improve system)
• A/B testing framework (test prompts, models, approaches at scale)

Observability:

• Agent tracing (debug decision chains)
• Cost monitoring (LLM token usage per query, per use case at 500K+ scale)
• Performance dashboards (query latency, success rates, error tracking)

Tools: Arize platform (already set up for Amplify), adapt Marketing Agent eval patterns

Why critical: Learn fast, iterate quickly, hit quality bar from launch. Without evals, flying blind with burned team.

Infrastructure Principles

1. Separate data from intelligence

• Customer 360 Platform (data layer): Data360 or in-house Snowflake
• Intelligence capabilities (AI layer): We build Klaviyo-specific, regardless of data platform choice

2. Buy infrastructure, build intelligence

• Agent hosting/orchestration: Likely Actively (accelerator from 0 to 1M agents)
• Intelligence models: We build (our competitive moat)

3. Coordinate with in-progress work

• CMR building canonical IDs (use for identity)
• Data360 or Snowflake Customer 360 building (coordinate on requirements)
• Gong, Enrichment integrations in progress (leverage, don't duplicate)

4. Validate before committing

• Test both data platform options (performance, cost, API quality)
• POC Actively (prove custom intelligence development works)
• Decide with data, not hypotheses

5. Keep intelligence portable

• Intelligence layer works regardless of Customer 360 Platform choice
• Switching data platforms doesn't require rebuilding intelligence
• Switching agent infrastructure doesn't lose intelligence logic

Section 5: Current State, Dependencies & Critical Decisions

This section grounds the infrastructure discussion in reality—what exists today, what's being built by other teams, what we must build ourselves, and what critical decisions unlock the roadmap.

Foundational Work in Progress (Q1-Q2 2026)

Project Hermes (Launching Jan 15):

• Territory validation ("Can I work this account?"), customer checking, deduplication
• Enrichment via Clay waterfall (firmographics, technographics)
• Basic fit scoring capability (in development)
• Amplify leverage: Integrate immediately, extend for continuous monitoring vs point-in-time

Enrichment Service (Launching Jan 15):

• Middleware approach (Workato + Clay handle provider waterfall)
• Single source of truth per data point (vs 5 provider redundancy creating "Which revenue is correct?" confusion)
• Amplify leverage: Build intelligence on clean data, leverage for accuracy baseline

Customer Master Registry (Q1-Q2 2026, GrowthArc building):

• Canonical customer IDs (one legal entity = one account)
• Parent-child hierarchy (ultimate parent → subsidiaries)
• Master linking table (Registry ID connects all system IDs)
• All source data already syncing to Snowflake (Salesforce, Zendesk, product usage, NetSuite)
• Amplify leverage: Critical dependency for Customer 360—provides identity foundation for either data platform option

Data Platform (Q1-Q2 2026):

• Two paths under consideration:
  • Data360: Systems team considering for AgentForce (clean APIs, business user interface, cross-functional access)
  • In-house Snowflake Customer 360: Building full capabilities in-house (more control, lower cost, but need to build interface layer)
• Both leverage CMR for canonical IDs and identity resolution
• Amplify dependency: Need one path ready for building advanced intelligence capabilities (CMR is strict dependency, Customer 360 increases velocity)

Gong Integration (In Progress, Systems Team):

• Bringing transcripts into data platform
• MEDDPICC extraction planned via Workato
• Amplify leverage: Build on completed integration work, extend with broader intelligence

Timeline insight: All foundational pieces launching Q1-Q2 2026—coordination opportunity if we push for what Amplify needs.

Critical Dependencies

Customer 360 Platform (either Data360 or in-house Snowflake):

• Needed before: Building intelligence capabilities that synthesize across systems--particularly non-sales data (e.g., product usage, support info)
• Why critical: Enables clean consumption of cross-enterprise data, supports unified account memory, avoids point-to-point integration mess
• Status: Building Q1-Q2, path TBD

CMR Canonical IDs:

• Needed before: Identity resolution, deduplication, hierarchy management
• Why critical: Solves 70% unusable territory problem, parent-child confusion, account ownership conflicts
• Status: Q1-Q2 2026

Agent Infrastructure:

• Needed before: Hosting any intelligence capabilities at scale (1M accounts)
• Why critical: Can't run stateful agents without orchestration, context management, hosting environment
• Status: Actively POC validating vs custom build required

Data Quality Baseline:

• Needed before: Building on unreliable data (garbage in → garbage out)
• Why critical: 98-99% accuracy requirement—burned team won't adopt at 95%
• Status: Cleanup prioritized Q1 2026 via CMR + Enrichment Service

What Amplify Must Build

Regardless of vendor leverage, we own:

1. Klaviyo-Specific Intelligence Models

• ICP scoring (trained on our conversion patterns, not generic vendor models)
• Win probability (our deal outcomes, our MEDDPICC patterns)
• Competitive positioning (Klaviyo vs Braze/Mailchimp/Iterable—our angles)
• Expansion logic (our product usage → upsell patterns)
• Next action recommendations (our workflow logic, our best practices)

2. Custom Integrations & Tracking

• LinkedIn monitoring (champion job changes, company posts, activity signals)
• Email/calendar tracking (engagement metrics, response patterns, meeting attendance)
• News/funding APIs (external timing signals)
• Note: Most source integrations already exist (Salesforce, Zendesk, product usage → Snowflake), Glean MCP potential shortcut to others

3. Command Centers (Tailored UX for Our Workflows)

• Persona-specific interfaces (BDR, AE, Manager, CG AE)
• Approval & outreach mgmt. workflows (review agent recommendations)
• Mobile access, Slack integration

4. Evaluation Infrastructure (Critical for Quality Bar)

• Eval datasets (ground truth for all predictions)
• Accuracy tracking (predicted vs actual outcomes)
• Human feedback loops (rep corrections improve system)
• A/B testing framework (test prompts, models at scale)
• Cost monitoring (LLM spend at 500K-1.5M scale)

Team requirement:

• With maximum vendor leverage (Actively + Data360 or Snowflake Customer 360): 6-8 engineers
• Full custom build (no vendor accelerators): 12-15 engineers

Critical Decisions

These decisions unlock the roadmap. Without them resolved, can't build intelligence or ship capabilities.

Decision 1: Customer 360 Platform

Validate both options:

Data360:

• Query performance at 500K-1.5M scale with high agent query volume
• Total cost model (licensing + usage at our scale)
• API quality for building AI applications (schemas, metadata, clean data extraction)
• Business user interface for cross-functional access (sales, support, success)
• Timeline alignment with CMR delivery

In-House Snowflake Customer 360:

• GTS Data team capacity and timeline
• What interface layer would they provide for AI application development?
• What business user access would they build? Should Amplify own interface layer?
• Query performance with in-house implementation
• Total build effort required (from GTS Data and Amplify)

Both viable. Key tradeoffs: speed vs control, vendor APIs vs custom build, cost vs flexibility.

Decision 2: Agent Infrastructure

Validate Actively (POC in Progress):

• Can we build Klaviyo-specific intelligence on their SDK? (POC Track 2: extensibility validation)
• BDR value delivered? (POC Track 1: 70%+ acceptance, conversion ≥ manual, productivity lift)
• Cost at enterprise scale justified by ROI? (low seven figures annually)

If POC successful: Actively accelerates from "can't host agent" to "1M agents at scale" (2-3 engineers, faster timeline)

If POC unsuccessful: Custom LangGraph + AWS (3-5 engineers, longer timeline, more control)

Decision 3: Integration & Coordination Strategy

Coordinate with Systems team:

• Who owns long-term maintenance of integrations? (Amplify, Systems, or partnership)
• How do we influence Data360/Snowflake Customer 360 design for intelligence needs?
• Integration with CMR—ensure canonical IDs work with chosen data platform

Pattern: Maximize native integrations (where they exist), build custom only where needed (LinkedIn, email/calendar, news)

Coordination Opportunity

The alignment: Hermes, Enrichment Service, CMR, and Data Platform all launching Q1-Q2 2026. If we coordinate actively—submit use cases to CMR team, push Data360 configuration (or influence Snowflake Customer 360 DB design), partner on integration ownership—timeline is achievable.

If we wait passively for foundational work to complete without influencing it for our needs, we risk delays and solutions that don't support intelligence requirements.

Action required: Push for what Amplify needs from in-progress work. Don't assume it happens automatically. The systems team feedback was clear—they want PM-led coordination and clear use cases.

Section 6: Roadmap & Execution Plan

Roadmap at a Glance

Phased Execution

Phase 1: Pre-POC Setup

Timing: Before Actively POC starts

Purpose: Establish baseline, build foundational infrastructure, prepare team

What we're building:

Time Study (Critical Baseline):

• Detailed time tracking across BDRs, AEs, Managers, CG AEs
• Measure where time actually goes today (validates 80% pipeline generation hypothesis)
• Output: Clear ROI targets per capability ("saves X min/week per rep")
• Informs post-POC priorities (what capabilities deliver most value)
• Why critical: Need "before" measurement to prove "after" improvements

Eval Infrastructure:

• Configure Arize for agent performance tracking
• Define metrics: acceptance rate, conversion rate, time saved, accuracy, adoption
• Set up data collection (track every POC interaction)
• Why critical: Measure POC from Day 1—evals aren't infrastructure to build later, they're how we validate success

Data Quality Baseline:

• Leverage Enrichment Service (Jan 15 launch)
• CMR deduplication beginning (Q1 start)
• Validate baseline accuracy of Salesforce, enrichment data
• Target: Clean data before POC (garbage in → garbage out)

Team Capability Building:

• Build simple agent pilot (SQL inbound processing): Learn end-to-end agent development, practice eval setup, build integration muscle before Actively POC
• Agent development fundamentals (if building on Actively SDK)
• Learn eval methodology on real agent (not just theory)
• Define success criteria with BDR team/manager

Deliverable: Ready to run POC with measurement infrastructure in place, baseline documented, team has agent development experience

Phase 2: Actively POC (BDR Validation)

Timing: Jan-Feb (3-4 weeks post-procurement approval)

Purpose: Validate BDR value + prove API extensibility for future use cases

What we're building:

4 Core BDR Capabilities (via Actively):

1. Account Prioritization: Daily ranked list (top K accounts) combining ICP fit, intent, timing
2. Account Research: Complete briefs (company, tech stack, pain points, buying committee, competitive context)
3. Contact Prioritization: Ranked contacts within accounts with reasoning
4. AI-Drafted Outreach: Personalized emails + call talk tracks, pre-written for Gong Engage sequences

Delivery Mechanism (Actively Agent Inbox):

• Push: Daily delivery to 10-12 pilot BDRs via Slack
• Pull: On-demand queries for any account
• One-click orchestration: Add to Gong Engage with full sequence

Simple Dashboard (Day 1):

• Vercel dashboard surfacing Actively intelligence
• Track acceptance/rejection of recommendations
• Proves API extensibility (not locked into Actively UX)
• Iterate based on BDR feedback

Validation Metrics (Evals Running):

• Acceptance rate: Do BDRs trust recommendations? (target: 70%+)
• Conversion rate: Agent-selected accounts ≥ manual selection
• Productivity lift: Time saved (research 30min → <5min, prioritization 1hr → 5min)
• Go/No-Go: If metrics hit, proceed to scale. If not, reassess.

Deliverable: Validated BDR value, API extensibility proven, decision on Actively vs custom infrastructure

Dependencies: Actively POC approved, pilot BDR team identified, baseline measurement complete

Phase 3: BDR Scale (Roll Out Fast)

Timing: Q2-Q3 (assuming POC successful)

Purpose: Get validated capabilities to all BDRs quickly—don't wait if it works

What we're building:

Scale Validated Capabilities:

• Account prioritization across all BDR segments (SMB, Mid-Market, Enterprise)
• Research brief generation for all accounts
• Contact prioritization and buying committee intelligence
• AI-drafted outreach with rep voice learning
• Roll to all BDRs (200-300), not slow 10 → 50 progression

Dashboard Iteration:

• Refine based on POC feedback
• Add features reps request
• Polish UX (still lightweight, not over-engineered)
• Mobile-responsive

Continuous Improvement Loops:

• Evals operational: track what works, what doesn't
• Model improvements: refine ICP scoring, prioritization weights based on conversion data
• Rep feedback integration: iterate on research quality, email tone, contact reasoning
• Platform gets smarter as more reps use it

Hermes & SQL Triage Integration:

• Integrate Hermes territory validation (Jan 15 launch)
• Inbound lead processing automation
• Leverage, don't duplicate

Deliverable: BDR transformation complete—all BDRs using Intelligence Platform daily, major efficiency gains proven at scale

Dependencies: POC successful (70%+ acceptance validated), CMR ready for identity resolution, data platform accessible (Data360 or Snowflake Customer 360)

Phase 4: AE Expansion & Automation

Timing: Q3-Q4

Purpose: Expand beyond pipeline generation to deal progression—full pipeline → close coverage

What we're building:

AE Capabilities:

• MEDDPICC Auto-Extraction (eliminate 60 min/day CRM time)
• Deal Health Monitoring (7-14 day early warning)
• Win Probability (accurate forecasting)
• Automatic CRM Updates (stage, next steps, activity logging)
• Meeting Coordination, Follow-Up Automation
• Proposal Generation

Automation Infrastructure:

• CRM write-back (Salesforce API or Workato)
• Email/calendar APIs (engagement tracking, meeting scheduling)
• Workflow orchestration (SA coordination, demo prep)

Custom Integrations:

• LinkedIn monitoring (champion job changes, company posts)
• Email/calendar tracking (response patterns, meeting attendance)
• News/funding APIs (timing signals)

Dashboard Expansion:

• AE view (deal priorities, complete context, health alerts)
• Approval workflows for automation

Deliverable: Full GTM org coverage (BDRs + AEs), zero CRM time, automated workflows operational

Dependencies: Data platform ready (Gong transcripts, Salesforce data integrated), BDR foundation proven

Phase 5: Manager, CG AE & Continuous Improvement

Timing: H1 2027

Purpose: Complete lifecycle intelligence—team oversight + post-sale capabilities + organizational learning

What we're building:

Manager Capabilities:

• Manager Dashboard (team performance with root cause analysis)
• Performance Analytics (replicate top performers—40% self-sourced vs team 20%)
• Coaching Recommendations (data-driven, specific per rep)
• Replaces 700 individual reports

CG AE Capabilities:

• Expansion Propensity (usage patterns → upsell readiness)
• Churn Risk Prediction (60-90 day early warning with root cause)
• Surfaces invisible context (support ticket history Maddie can't see today)

Continuous Learning:

• Email response pattern learning (what messaging works by segment)
• Call timing optimization (codify Keegan's West Coast 5-8 PM discovery)
• Deal progression playbooks (extract from won/lost patterns)
• ICP model evolution (refine based on conversion data)

AI Assistant:

• Natural language queries ("Which accounts should I prioritize?")
• Addresses reporting vacuum (conversational vs 700 reports)

Deliverable: Full lifecycle intelligence operational, organizational learning compounding, complete transformation (600 reps)

Dependencies: Product data integrated (Snowflake usage, Zendesk support), sufficient interaction volume for pattern learning

Dependencies & Critical Path

Infrastructure enables capabilities:

Can start without data platform (Pre-POC):

• Eval infrastructure setup
• Team capability building
• Data quality baseline

Needs data platform (Q1-Q2):

• Intelligence capabilities (ICP scoring, account research, MEDDPICC extraction)
• Can't synthesize across systems without unified data
• Blocker: CMR + Data360/Snowflake Customer 360 must be ready

Needs Actively validated (Jan-Feb POC):

• BDR scale (if POC successful, roll fast)
• Agent hosting at scale
• Blocker: If POC fails, 6-12 month delay for custom infrastructure

Needs product data (later):

• CG AE capabilities (expansion, churn)
• Manager advanced analytics
• Blocker: Product usage integration (Snowflake → data platform)

Faster if coordinated:

• Hermes integration (Jan 15—immediate value)
• Gong integration (Systems team—leverage vs rebuild)
• Enrichment Service (Jan 15—clean data foundation)

Scaling Philosophy

Not: Slow linear progression (10 → 50 → 100 → 300)

Instead: Validate → Scale fast → Expand personas

Validate (POC): Prove it works with 10-12 BDRs
Scale (if validated): Roll to all BDRs + ENTR/SMB AEs quickly (200-300 in Q2-Q3)—don't wait if it works
Expand (once proven): Add AEs, then Managers, then CG AEs—each builds on previous foundation

Rationale: If capabilities deliver 70%+ efficiency and BDRs adopt, organizational imperative to roll out fast. Slow rollout leaves value on the table.

Appendix A: Discovery Research - Full Analysis

Period: November 25 - December 11, 2025
Methodology: 11 stakeholder interviews + 4 shadow sessions (Enterprise BDR, SMB BDRs × 2, ENTR AE)
Coverage: Systems teams (GTM, BPA, Engineering), Sales leadership (BDR Directors, Enablement, Onboarding), Frontline reps (BDRs, AEs across all segments), Cross-functional (CFO CoS, Data/MDM teams)

Introduction: What We Did

Over three weeks, conducted comprehensive discovery to understand current sales operations, identify efficiency bottlenecks, and validate solution direction before building. Approach combined structured interviews with systems teams and sales leadership, frontline conversations with BDR directors and reps, and hands-on observation through shadow sessions. Goal was understanding actual workflow, not just stated problems—what do reps do all day, where does time go, what creates friction, what would actually help.

Discovery activities:

• Week 1 (Nov 25-Dec 1): Systems teams (Joe Rodden - GTM Systems, Glenn Vanderlaan - BPA, Chad Marxen - Solutions Architect), Engineering (Noah Mormino), Cross-functional (Griffin Dowdy - CFO CoS, Indrajit - MDM lead)
• Week 2 (Dec 2-3): Sales/Enablement (Morgan Jacobson - Head of Enablement, Eric Gallito - Onboarding PM/former AE), Project Hermes team (Justin Hiatt, Jake Rainess, Josh Craig)
• Week 3 (Dec 9-11): BDR leadership (Emma Dwinells - Mid-Market/Enterprise Director, Alexis Caso - SMB/Entrepreneur Director), Frontline shadows (Charlotte Huang - Enterprise BDR, Keegan Minahan - SMB BDR, Maddie White - ENTR AE, Zoe Verdone - SMB BDR)

What made this effective: Combined top-down (leadership) and bottom-up (frontline) perspectives. Validated patterns across multiple sources. Observed actual workflow (shadow) vs just hearing about it. Asked "show me" not just "tell me."

Executive Summary

Discovery revealed that pipeline generation consumes 80% of rep time and is the core driver of sales efficiency—not closing skills, not deal management, but finding and engaging the right prospects. The pipeline generation process breaks into three core activities (prioritize accounts, identify contacts, personalize outreach), and each activity suffers from systematic problems: tool sprawl (15+ systems requiring manual synthesis), data quality issues (trust broken from repeated bad experiences), and missing intelligence (no scoring, no prioritization, no synthesis across signals).

Critically, shadow session validated the solution: the BDR's unprompted "magic wand" request—"AI gives me 10 accounts prioritized daily, which people have best chance responding, what to message them"—exactly describes the Intelligence Platform in Section 1. Her three questions (accounts, people, messaging) map to core capabilities. Her workflow pain points map to roadmap priorities. She's already envisioning what we're planning to build.

The organizational landscape: Systems teams are building foundational pieces (CMR for identity in Snowflake, Data360 or in-house Customer 360 for interface, Hermes for BDR workability, Enrichment Service for data quality). All launching Q1-Q2 2026, aligning with Amplify pilot timing. The sales/systems relationship is strained, creating governance gap that Amplify fills with PM-led coordination. Opportunity: Extend work in progress vs duplicate, fix relationship, deliver value fast.

Confidence level: High. Findings consistent across 11 stakeholders. Solution validated by end users. Priorities clear (pipeline generation first). Architecture aligns with in-progress work (CMR + data platform). Quick wins identified. This research de-risks execution.

Pipeline Generation: Core Driver of Sales Efficiency

Why Pipeline Generation Drives Efficiency

Pipeline generation emerged as the primary constraint across all stakeholder interviews—sales leadership, enablement, BDRs, and former AEs:

"Pipeline creation is 80% of the job. Two modes for reps: creating dollars (pipeline generation—the bottleneck) and closing dollars (deal progression—secondary). More pipeline → hit targets → everyone happy."

"Pipeline is the biggest issue. Going into this year, it's actually kind of impossible to hit quota for most reps because there's not enough demand."

"Prospecting and pipeline generation is the bottleneck. Should absolutely go after first. Number of accounts we can go after is biggest driver of efficiency."

"Discovery and demo optimization won't have tremendous impact. Better metric: Average rep pipeline creation $5K → $15K/month—way more impressive."

The math that makes pipeline critical:

• Traditional coverage requirement: 3-4x pipeline to quota
• Example: $10K monthly quota with 80% win rate → need to generate $12.5K+ weighted pipeline monthly
• Close rates are relatively fixed (can improve 5-10% with enablement, but mostly is what it is)
• Revenue calls per day are limited (AEs: 2-3 discovery/demo calls, couple hours max)
• The denominator problem: Can't significantly increase calls per day, so need more total opportunities to work

Not just an efficiency problem—it's a demand generation problem. Insufficient pipeline makes quota unattainable regardless of how well reps close.

Time allocation reality from stakeholder interviews:

• 80%: Finding accounts, researching prospects, booking meetings (pipeline generation)
• 20%: Discovery, demo, closing (actual selling)

Quantified from project teams:
"BDRs spend 30-50% of their time info gathering before they can even prospect effectively. This isn't inefficiency—it's necessary for quality prospecting in our current environment."

Focusing on pipeline generation = focusing on 80% of rep time and the actual constraint on revenue.

The Pipeline Generation Process (Three Core Activities)

Framework validated across BDR directors and frontline reps. Every BDR/AE asks three questions daily:

Activity 1: Prioritizing Accounts - "What accounts should I go after today?"
Activity 2: Identifying Contacts - "Who are the right people at those accounts?"
Activity 3: Personalizing Outreach - "What should I say to them?"

Each activity has specific tools, specific pain points, and specific time consumption. Understanding problems within each activity reveals where to build solutions.

Activity 1: Account Prioritization

Current Process (Observed During Shadow + Validated)

Morning routine (before any outreach):

Check 6sense for intent signals (15 minutes)

• Platform shows ~20-25 "hot" accounts (recent website activity, research behavior)
• Assessment from reps: "Not always totally accurate, gives ballpark"
• Manual filtering required: Many shown are customers, non-tech-fit businesses, or universities
• Outcome: Find 1-2 accounts actually worth adding to outreach list

Review Salesforce territory list (15 minutes)

• Observed territory: 783 total accounts
• After manual categorization project: ~200 workable (attempting + prospect)
• Rest categorized as: customers (unmarked), invalid URLs, duplicates, B2B non-fits, no contacts found
• Review attempting list, decide priorities based on closed-loss timing, recent activity, deal size potential

Check for trigger events (10-15 minutes)

• Job changes: People who recently changed roles (LinkedIn - manual checking)
• Events: Upcoming conferences/dinners to invite prospects
• Closed-loss timing: Contracts potentially ending (manual tracking)

Decide today's focus (5-10 minutes)

• From ~200 workable accounts, pick 10-20 to focus on
• Based on: Intent signals + timing + deal size + experience-based judgment
• No systematic scoring methodology

Total morning prioritization time: 45-60 minutes before first outreach

The Prioritization Problems

Problem 1: No systematic multi-factor scoring

Currently deciding based on individual factors checked separately:

• Intent signals (6sense - requires manual filtering)
• Closed-loss timing (manual tracking in Salesforce)
• Deal size potential (firmographic data from multiple sources)
• Tech stack fit (Shopify Plus vs other platforms)
• Territory workability (manually categorized)

Missing: Combined intelligence synthesizing all factors into single score

• ICP fit (firmographics + technographics + behavioral signals)
• Intent signals (6sense + website activity + LinkedIn + news)
• Timing triggers (contract expiration, budget cycles, stakeholder changes)
  • Critical gap validated across shadows: Contract renewal data exists but not surfaced
  • Zoe (SMB BDR): "Find an account that says their contract ends December 2025 and I'm like, how did I not know that? Why was that not on some report for me? That was pure luck."
  • Seasonal timing also manual: December = target Mailchimp/Shopify for quick wins before holidays
• Past engagement (response history, relationship strength)

Territory assignment challenges:

• Zip code-based allocation creates inequity
• Revenue alone inadequate ($50M company with 5K contacts = lower value than small company with 1M subscribers)
• Need: Expected value model (deal size × conversion probability)
• Need: Multi-factor scoring combining 30+ signals

Problem 2: Firmographic research requires manual aggregation

Multiple tools and Chrome extensions used during prioritization:

• Similar Web (Chrome extension): Traffic data, growth rates, deal size indication
• BuiltWith (Chrome extension): Tech stack detection
  • Key signal: Using 4+ MarTech solutions indicates marketing investment
  • Platform matters: Shopify Plus vs WooCommerce vs Magento
• Clearbit (Salesforce fields): Employee count, revenue estimates, industry
  • Issue: Multiple provider fields create clutter (D&B revenue vs Clearbit revenue vs Similar Web revenue)
• ZoomInfo: Funding status, headquarters, growth indicators
• 6sense: ICP fit score, profile strength ratings
  • Assessment: "Gives ballpark" but not definitive

Current workflow: Open 3-4 sources, manually review each, synthesize mentally, decide if worth pursuing

Problem 3: Territory workability reality

Observed during interviews:

• Mid-Market BDR territory: 500 accounts → only 150 workable (30%)
• Enterprise BDR territory: 783 accounts → ~200 workable after categorization
• 70% of assigned territory is unusable

What's in the unusable portion:

• Customers not marked in system
• Invalid URLs (sites that never worked)
• Duplicates
• B2B companies (banks, manufacturers - not target market)
• No contacts findable
• Parent-child confusion

Manual categorization required:

• Described as "such a force"
• Every account reviewed by BDRs
• Example non-fit: Bank of America ("banks don't sell products, compliance issues, not B2C")

Geographic challenges with zip code territories:

• E-commerce doesn't map to geography naturally
• NYC dense with accounts, outlying areas sparse
• Some pairings cover disparate regions (North Carolina + Minnesota)

Problem 4: Prioritization is static, not dynamic

Morning decision creates fixed list for the day:

• No reprioritization if signals change during day
• Intent spike at 2 PM doesn't surface
• Champion response doesn't trigger next steps automatically

Contrast with needed state: Continuous reprioritization based on real-time signals

Activity 2: Contact Identification

Current Process (Demonstrated During Shadow)

Find broad contact list in LinkedIn Sales Navigator (5 minutes)

• Search target company
• Apply Boolean filter on job titles (marketing, analytics, operations, etc.)
  • Described as "gold standard, taught by manager on first day"
  • LinkedIn learns patterns over time
• Result: 30-40 people with relevant titles
• Save to lead list (organized by account, event, or campaign)

Prioritize within contact list (10-15 minutes)

Manual analysis per person:

• Review title: Senior enough for decision authority?
• Check background: Hands-on role or strategic? (hands-on users care about platform)
• Review LinkedIn activity: Engaged? Posts about relevant pain points?
• Assess function relevance: Marketing ops vs content vs loyalty (ops most directly relevant)
• Note geography: International considerations for timing, language, approach
• Cross-reference Salesforce Activity: Any past touches? How did they respond?

Questions being answered manually:

• Is this person a decision maker? (budget authority)
• Have they talked to us before? (warm vs cold outreach)
• Do they know their domain? (technical competence)
• Do they work in platforms daily? (hands-on vs oversight)

Time investment: 10-15 minutes to select 8 priority contacts from initial 40

This represents hidden time sink: Appears like simple selection, actually requires significant analysis

Export contacts via ZoomInfo extension (2-3 minutes)

• Select decided contacts in LinkedIn
• ZoomInfo extension: Retrieves contact information (email, mobile, title verification)
• Exports simultaneously to Gong Engage (adds to outreach flow) and Salesforce (creates contact records)
• Bridge function: ZoomInfo connects LinkedIn selection to execution systems

Recent operational change noted:

• Now requires email present to add contacts to Gong
• Previously: Added all contacts, inferred emails from company domain
• Current: Missing email requires manual Salesforce contact creation first
• Described as "longer process, people get lost in workflow"

Total time per account: 17-23 minutes for contact identification

Contact Identification Problems

Problem 1: Broad filtering easy, prioritization hard

Boolean search produces 30-40 relevant people quickly. Ranking them requires pattern recognition that early-career BDRs don't have. Experienced reps use 6 months+ of mental models to assess decision authority, platform engagement, and champion potential. New BDRs take longer, make lower quality selections, experience more trial and error.

Problem 2: Contact data quality varies

Enterprise contacts generally accurate (high-level people easier to locate). SMB contacts have quality issues. Validation problems include wrong phone numbers, bounced emails, and LinkedIn data staleness (showing previous company from 6 months prior). Outdated LinkedIn data causes incorrect account routing.

Problem 3: Past engagement context is buried

To determine if contact was previously reached:

• Navigate to Account page
• Click Activity tab
• Scroll through all activities
• Look for specific contact name
• Check response status

Time-consuming and easy to miss. Information not surfaced on account homepage.

Problem 4: Deep account penetration not systematic

Current reality: BDRs contact 2-3 people per account on average. Ideal state (from BDR leadership): 50 prospects per week per account, cascading through entire team (marketing → IT → data → operations → c-suite). Can't spam all 100 at once (inbox blocking), but should systematically add weekly.

Example impact: Southwest account has 239 identified prospects, only 2-3 contacted. Same account, 100x untapped potential.

Problem 5: LinkedIn workflow has constraints

InMail credits: 50 per month (carries over if unused). Cannot message all people in outreach flow. Requires strategic selection of 3-4 highest-priority contacts. Others receive connection request only, hoping for acceptance. Results in LinkedIn activity being out of sync with Gong flow task management.

Activity 3: Outreach Personalization

Current Process (Demonstrated During Shadow)

Research for personalization (10-15 minutes per account, first time)

Personal information gathering from LinkedIn:

• Career background and experience (progression, past roles)
• Bio section: Personal hooks and interests
  • Example: Bio mentions "love working in e-commerce" - reference in outreach
  • Example: Previously worked at existing Klaviyo customer - mention shared connection
• Recent activity: Posts, comments revealing pain points or priorities
• Education and certifications (conversation starters)

NOT heavily researched during outreach personalization:

• Firmographic data (revenue, employee count) - already reviewed during account prioritization
• Tech stack details - already factored into account selection
• Traffic/growth data - already used in deal size estimation

Company context from Salesforce:

• Navigate to Opportunities tab
• Find closed-loss opportunities if they exist
• Read notes: timing, involved parties, loss reason, deal size
• Often incomplete (depends on AE note-taking)
• Copy relevant context for email reference

Draft email using AI agent (5-10 minutes initially, then iterations)

Process:

1. Paste LinkedIn profile URL into teammate-built ChatGPT agent
2. Add closed-loss context if available
3. Add standard template
4. Receive draft email

Assessment from shadow observation:
"Way better than I could write, lot quicker. BUT still not amazing. First message wasn't great. Kind of cheesy, don't love this."

Iteration cycle observed:

1. Review first draft (30 seconds)
2. Identify issues: generic voice, weak call-to-action
3. Request refinements: "Rewrite this, make format more like 'hospitality teams we work with see X'"
4. Review again: "Better call to action needed"
5. Repeat 2-3 more times
6. Total: 10-15 minutes to reach sendable quality

Why current AI requires iterations:

• Generic voice (team-wide agent can't personalize to individual style)
• Formulaic structure (recognizable patterns)
• Weak calls-to-action
• Insufficient account-specific context
• Personal agents work better (trained on individual writing style over time)

Format and send (2-3 minutes)

• Copy to Gong Engage
• Remove AI-looking formatting
• Adjust subject line (observed preference: lowercase, brief - "quick chat")
• Final typo check
• Send manually (not automated due to quality concerns)

Total time per first-touch email: 20-25 minutes with AI assistance, 30-40 minutes without
Follow-up emails: Faster (5-10 minutes - brief updates, reiteration)

Knowledge capture gap:

• Google Sheet "account map" maintained with AE for priority accounts
• Should contain: research notes, tech stack, opportunities, messaging angles
• Reality: "Takes time nobody wants to spend, not always up-to-date, lost if person quits"
• Most intelligence remains in rep's head or scattered notes

From shadow observation:
"My notes app is so disorganized—million random things I think of. Notes are in my head or scattered. If I quit, lost forever."

Outreach Personalization Problems

Problem 1: Research aggregation is manual

Information needed for effective personalization:

• Personal context (bio, background, past companies, interests)
• Company pain points (LinkedIn posts, industry trends)
• Past Klaviyo engagement (buried in Salesforce Activity)
• Closed-loss history (what happened, why, when)
• Competitive intelligence (current tools, known frustrations)
• Similar customer proof points (relevant case studies)

Currently: Rep manually aggregates from 5+ sources over 10-15 minutes

Missing: Automated synthesis into single brief

Problem 2: AI email quality requires iteration

Team-built agents provide starting point but don't deliver sendable quality:

• First drafts described as "kind of cheesy"
• Require 3-5 refinement cycles
• Still takes 10-15 minutes even with AI
• Personal agents perform better but require individual training

Gap between current ("better than I could write") and needed ("perfect first draft, minimal edits")

Target: 20-25 minutes → 5 minutes for first-touch emails

Problem 3: Knowledge loss is systematic

Research happens but isn't captured:

• LinkedIn insights remain in rep's memory
• Account intelligence not systematized
• Messaging strategies not documented
• Every departure = permanent intelligence loss

Wish from interviews:
"As you do research—copy LinkedIn, write email—information gets saved automatically. Or AI scrapes internet weekly, auto-updates account map."

Cross-Cutting Issues Affecting All Activities

Tool Sprawl & Integration Gaps

The complete tool landscape across three activities:

Account prioritization phase:

• 6sense (intent signals)
• Salesforce (territory, categorization, opportunities)
• Similar Web (traffic - Chrome extension)
• BuiltWith (tech stack - Chrome extension)
• Clearbit (firmographics in Salesforce)
• ZoomInfo (funding, growth)
• Google (validation, news)

Contact identification phase:

• LinkedIn Sales Navigator (search, Boolean filtering)
• Salesforce Activity (past engagement)
• ZoomInfo extension (contact info retrieval, export bridge)

Outreach personalization phase:

• LinkedIn (profile research)
• Salesforce Opportunities (closed-loss context)
• Salesforce Activity (call recordings via Gong links)
• Google Sheets (account map - manual, AE coordination)
• Notes app (scattered observations)
• ChatGPT / Custom GPT agents (email drafting)
• Gong Engage (flow management, task queue, sending)
• Gmail (full email thread view, manual bumping of old conversations)
• Calendar (manual follow-up reminders - "4 days later, follow up with this person")
• Spreadsheets with AEs (tier 1 targets, notes, coordination - not standardized)

Total: 15+ distinct tools/systems used daily

Impact beyond time investment:

• Cognitive switching costs: Different interfaces, search patterns, authentication
• Information silos: 6sense intent not connected to Salesforce closed-loss context
• Manual synthesis required: Rep correlates signals mentally across disconnected systems
• Context loss: LinkedIn research not captured in Salesforce, insights lost
• Navigation tax: Knowing exact path to find information (Activity tab for calls, Opportunities for closed-loss)
• Follow-up chaos: Zoe (SMB BDR) works across Gong (tasks) + Gmail (thread view) + Calendar (reminders) to manage follow-ups
  • Dream feature: "I would do anything for a tab in Gong showing accounts by days since last touch: 1 day, 2 days, 3 days, 4 days, 5+ days"
  • Currently memory-dependent: "I remember because I have a good memory" - doesn't scale

Observation from shadow:
"We pride ourselves on being consolidated platform, but to work here we use 20 platforms. Should have platform like Klaviyo for our own process."

Alternative mentioned in interviews: BDR directors noted similarity to Nooks (prospecting platform providing single interface). Internal build potentially feasible given work already in progress.

Data Quality & Trust Crisis

State of data quality (consistent across interviews):

"We are behind where companies were five years ago. Our CRM is rampant with misinformation. To find a prince, you have to kiss 7 million frogs daily."

Context: Marketing automation platform launched only months ago. Previously had no systematic approach. Result: duplicates, incorrect enrichment, poor data quality pervasive.

Trust breakdown pattern:

• Tools deployed with inaccurate data
• Reps try once, encounter errors
• Permanent rejection: "Used it once, was wrong, never touched again"
• Cannot recover even after improvements

Example: Territory/segment identification tool rolled out. Data quality issues led to zero adoption despite potential utility.

Visibility gaps compounding trust issues:

• Maddie (ENTR AE): 75% of her book is churned/previous Klaviyo customers, but support ticket history completely invisible
• Quote: "Did they have the worst support experience ever? I didn't see that. I have no idea why they left."
• Critical context missing when re-engaging churned accounts

Comparison from other industries: Similar pattern with slide generation tools—terrible initially, improved over months, but impossible to recover adoption after bad first impression.

Specific data quality issues identified:

Field mapping errors (from systems team sprint reports):

• Lead → Contact conversion causes data loss
• "Ongoing issues" affecting data integrity
• Sync inconsistencies across systems

Country field problems:

• Names populating instead of codes
• Breaking enrichment and territory assignment
• Recently fixed but indicative of quality issues

Enrichment data clutter (historical problem being addressed):

• Point-to-point integrations from 5 providers created redundant fields
• Users confused: "Which revenue is correct? D&B vs Clearbit vs Similar Web?"
• New approach (launching Jan 15): Middleware handles waterfall, single answer returned
• Solves clutter but historical data quality issues persist

Hierarchy staleness:

• Acquisitions from year prior not updated in system
• Account ownership conflicts
• Parent-child relationships incorrect
• CMR project addresses (Q1-Q2 2026)

Deduplication backlog:

• Post-IPO 2-year cleanup effort
• Legacy PLG duplicates still present
• Ongoing challenge

Field overload:

• 500+ custom fields accumulated over 10 years
• Feedback from multiple stakeholders: "I don't know what half of this stuff is"
• Cleanup effort underway

What reps trust vs ignore:

Trusted data (actively used):

• Closed-loss opportunity notes (described as "gold standard" if complete)
• Activity logging (automatic, reliable)
• Account hierarchy (mostly accurate, validated externally when uncertain)
• MRR field (customer indicator, though fake trial accounts exist)

Untrusted data (ignored despite existence):

• Most enrichment fields (uncertainty about accuracy)
• Account ownership assignments (frequently incorrect)
• Contact routing (splits across wrong accounts)
• Anything not personally verified

The adoption requirement:

95% accurate = 5% wrong = Tools rejected by this team

Guidance from interviews:
"You're walking into this atmosphere. Only way forward: do really meaningful stuff that works from day one."

"Big proponent of AI but also big critic. Always verify everything. A lot of people put too much trust internally. If we said 'use X, Y, Z—this is source of truth'—that would go over better."

Need 98-99% accuracy for adoption. Burned team requires excellence from launch. Better narrow scope with high accuracy than broad capabilities with moderate quality.

Organizational & Governance Gaps

Sales/Systems Relationship Strain

Multiple stakeholders characterized the relationship:
"There is such an unhealthy feedback loop between sales and systems. It's the worst it's ever been."

Process failure documented:

1. Sales leader identifies needed change (Salesforce, Gong, tool modification)
2. Request submitted to sales ops
3. Sales ops creates ticket (requiring extensive back-and-forth on specifications)
4. Ticket enters systems team backlog
5. Systems team doesn't see request for 8 months (described as "graveyard")

Visibility breakdown:

• Sales unaware of systems priorities and roadmap
• Systems unaware of sales pain points and pipeline impact
• No shared understanding of how technical work connects to revenue
• Pipeline meetings reference "system issues" without clear ownership

Communication structure:

• Sales leaders → sales ops → systems (information lost in translation)
• No direct dialogue
• Workarounds: "sales ops goes back door to work with systems"

Resource and ownership challenges:

• Systems teams stretched across multiple areas
• Scattered responsibilities (marketing systems, GTM systems, sales ops, marketing ops)
• Unclear ownership: "Who owns this?" answered with "Not me, not me"
• No agility for rapid iteration

Tool acquisition dysfunction:

From systems team interviews:
"Operations says 'we bought this new tool, go implement it.' We're thinking 'we already have 7 tools that do that. Wish you brought us along for conversation.'"

Procurement decisions made without systems technical input. Results in capability duplication, implementation burden, and no coordination on overall tool strategy.

PM Governance Vacuum

Systems team feedback:
"Company's AI push good, but led to not as much PM support and governance. It's lot of people building—let's build first, figure out later if it's coherent."

Multiple teams building simultaneously:

• Hermes: BDR-built solution in Clay
• CMR: GrowthArc team in Snowflake
• Data360: Systems team for AgentForce
• Enrichment: Systems team for data quality
• Lead scoring: Marketing team
• Amplify: Intelligence platform

No coordination mechanism. Risks: duplication, incompatibility, wasted resources, solutions don't integrate.

Prioritization challenges:
"Historically, lot prioritized for operations instead of end users. Responded to who yells loudest instead of data-driven decisions."

• Operations demands get built
• End users remain unsatisfied
• Tension: Should systems push back without PM making calls?

What's needed from interviews:

• Data-driven product management
• Governance coordinating teams
• Systems involvement in acquisition
• Clear priorities with rationale

Engineering team perspective:
"Without strong product team, work lands on engineers. We're not product people. Work feels random—tackling initiatives other teams lack capacity for. Engineers filling product gap (not sustainable)."

What engineers want:

• Clear vision and rationale
• Structured roadmap in manageable pieces
• Stakeholder management and political cover
• Answers to open-ended questions
• Protection from scope creep

Reporting & Analytics Vacuum

Consistent feedback from BDR directors:
"700 individual reports per person. No core dashboards. Salesforce dashboards break when someone saves over them. Gong reporting trash."

Manager challenges:

• No standardization (every manager builds own in Excel or custom Salesforce reports)
• Individual proliferation (each BDR/AE maintains hundreds of personal Salesforce reports)
  • Described as discovering "closed-lost report from 8 months ago, should revisit"
  • No discoverability, maintenance, or relevance tracking
• Team dashboards fail: Created for group → someone overwrites → broken for everyone
• Tableau desired but never built (no bandwidth, though "saw beautiful dashboards" at prior companies)
• Gong analytics described as inadequate (can't extract meaningful health metrics)

Specific needs from BDR leadership:

Health metrics over time (not just current snapshot):

• Email open rates, connect rates, bounce rates, activity-to-meeting conversion
• Question they can't answer: "What was bounce rate in January vs now?"
• Granularity: Per individual AND per team
• Root cause analysis: "Connect rate 22% vs team 35%—WHY?"

Operational challenges for FY26:

• BDR:AE ratio changing to 1:4 (one BDR supporting four AEs)
• Need attribution: Where is pipeline coming from across four relationships?
• Need coverage visibility: Out of workable accounts, how many actively worked? Contacts per account?

Data exists but not accessible:

• Gong has all outreach activity
• Possibly synced to Snowflake (needs validation)
• Missing: Reporting layer, visualization, trend analysis

Impact beyond inconvenience:

• Can't coach with data (guessing vs root cause identification)
• Can't identify patterns (what messaging works, what doesn't)
• Can't replicate top performers (mentioned: one rep self-sources 40% pipeline vs team 20%—how?)
• Can't make informed strategy decisions (relying on intuition vs metrics)

Product Data Access Limitation

From systems interviews:
"If we can get APIs to product data, that'll be huge—it's a huge gap across the board. Everybody needs it but today you have to jump through hoops."

Cross-functional need for product usage data:

• Sales: Cross-sell signals (usage patterns indicating upsell readiness), account health scores
• Marketing: In-app behavior (feature usage, engagement patterns)
• CS: Expansion opportunities (hitting limits, requesting features), churn risk (declining usage)

Current state:

• Product usage data in Snowflake (separate from Salesforce sales data)
• Staffside internal tool has account details (BDRs noted "don't really utilize" - designed for support/CSM use)
• No easy sales access to product telemetry

Impact on intelligence capabilities:

• Cannot predict expansion without usage signals (hitting plan limits? using advanced features?)
• Cannot detect churn risk without engagement trends (declining sends, inactive users)
• Cannot identify whitespace without adoption data (missing products they should have)

Solution via CMR + Data360/Snowflake Customer 360:

• Product data already in Snowflake
• Data platform provides access (Zero Copy if Data360, custom interface if in-house)
• CMR canonical IDs link sales and product records
• Timeline: Q1-Q2 2026

What's Already Being Built (Coordinate vs Duplicate)

Project Hermes (Launching January 15, 2026):

• Solves: "Can I work this account?" (customer check, deduplication, parent-child validation)
• Enrichment via Clay waterfall (firmographics, technographics)
• Fit scoring capability (in progress, pending data access)
• MRR projection
• Custom UI (25% complete)
• Built by BDRs in Clay with LLM integration
• CEO-sponsored (three explicit goals provided)

Amplify coordination opportunity: Extend for continuous monitoring (vs point-in-time), integrate with broader platform, provide long-term productization

Enrichment Service (Launching January 15):

• Moves enrichment logic out of Salesforce (reduces UI clutter)
• Middleware approach: Workato + Clay handle provider waterfall
• API endpoint available for consumption
• Returns single answer per field (vs multiple provider redundancy)
• Phased rollout through March-April

Amplify coordination: Leverage for data quality foundation, build intelligence on clean data

Customer Master Registry (Q1-Q2 2026, GrowthArc):

• Canonical customer IDs (legal entity-based definition)
• Parent-child hierarchy (ultimate parent → ultimate child mapping)
• Master linking table (Registry ID connects all system IDs)
• All source data already syncing to Snowflake (Salesforce, Zendesk, usage, NetSuite)
• Provides: Query with one ID, retrieve all data across systems (eliminates point-to-point integration complexity)

Amplify coordination: Use as identity foundation, submit use cases to influence design, test with staging instances

Data360 / Salesforce Data Cloud (In Progress, Systems Team):

• Zero Copy integration with Snowflake (all data accessible without ETL)
• Vector search capability
• Open APIs for custom intelligence
• AgentForce dependency (already committed)
• Coordinating with CMR team on canonical IDs

Amplify coordination: Validate query performance, configure for intelligence use, leverage as interface layer

Gong → Data Platform Integration (Systems Team):

• Bringing transcripts into data platform (Data360 or in-house Snowflake)
• Enables vector search (vs repeated API calls)
• MEDDPICC extraction planned via Workato
• In active development

Amplify coordination: Leverage completed work, extend with broader intelligence capabilities

Continuous TAM Discovery Opportunity:

• BI team conducts periodic uploads (platform-specific cuts: Shopify, BigCommerce, Magento)
• Historically manual from StoreLeads
• Expansion opportunity: New verticals (restaurants, spas, services)
• Automation potential: Continuous discovery vs periodic manual

Validation of Solution Direction

Shadow Session Validates Vision

The BDR's unprompted request when asked about ideal state:
"In an ideal world, every day I come in, AI gives me 10 accounts prioritized, which people have best chance responding, what to message them."

This precisely describes the Intelligence Platform in Section 1:

• Morning brief with prioritization (10 accounts "Act Today")
• Complete account context (3-minute review vs 30-minute manual research)
• Pre-drafted personalized emails
• Contact intelligence (best-response-probability contacts identified)

The three-question framework maps directly:

1. What accounts? → Account prioritization intelligence (P0 capability)
2. What people? → Contact prioritization intelligence (P0 capability, revealed as hidden time sink)
3. What to say? → AI-drafted messaging (P1 capability, quality bar is high)

Research Validates Priorities

Pipeline generation focus confirmed: Consistent across all stakeholders, quantified at 80% of rep time, identified as primary constraint

Three-activity framework validated: Observed in practice, time allocation documented, pain points specific to each activity

Tool consolidation critical: Not convenience but necessity (15+ systems creates information loss and cognitive burden)

Data quality prerequisite: Cannot build on unreliable foundation, cannot recover from bad user experience

Coordination opportunities clear: CMR, Data360/Snowflake Customer 360, Hermes, Enrichment all launching Q1-Q2 2026, aligning with Amplify pilot

This research de-risks execution: Building what users envision, solving experienced pain, coordinating with in-progress work, validated priorities.