Dugout
Spec

How Dugout is designed and rolled out

Three sections — architecture, rollout, and a companion system. Same page; jump links below.

Architecture

Four layers — built bottom-up. Each ships only if the layer below earns trust.

01
Data
Read-only ingestion.
  • Salesforce — accounts, opportunities, contacts, OCR, activities
  • Gong — call transcripts + risk markers
  • Outreach — sequence enrollment, email/call activity
  • Dock — buyer engagement on shared assets
  • Chili Piper — meeting booking events
  • ZoomInfo — account firmographics + buyer enrichment
02
Signal engine
Pure functions over CRM state.
  • Rules library — each rule = (id, severity, strategicPriority, evaluate fn)
  • Deal Health — compound state weighted by close-date proximity
  • Playbooks — multi-phase workflows attached to specific signals
03
Orchestration
Tasks: state, history, notes. Signals become work items, not alerts.
  • Reconciliation — signals ⇌ tasks every render
  • Lifecycle — open / done / snoozed / muted (with reason)
  • Notes — AE work notes + manager coaching notes
  • Auto-resolve — when a signal stops firing, the task closes itself + a toast surfaces it
04
AI synthesis
LLM only where it earns its keep.
  • Morning digest — Claude synthesizes the briefing from signals + transcripts
  • Sentiment + risk extraction on call transcripts (precomputed in seed)
  • Studio v2 (deferred) — NL → runnable rule

Design choices worth defending

Severity is the product

A morning digest with 30 line items is worse than no digest. Tier dictates routing. Routing is the trust contract.

Tasks, not alerts

State + history is what makes this an orchestration engine instead of a notification firehose. A signal you can't close is noise.

Read-only in v1

We read from the CRM. We don't write back until rules have earned trust. Bad writes are unrecoverable trust losses.

LLM second

Deterministic rules over structured data are cheaper, faster, testable. The LLM runs on synthesis tasks, not as a general agent.

The catalog (13 rules)

RuleSeverityPriority
Champion has left the company
LinkedIn signal detected the champion has changed jobs. This is the highest-leverage moment for intervention — the next 14 days determine whether the deal survives.
BLOCKINGP5 · Sales Motion Maturity
Champion silent for 7+ days
Champion hasn't replied, joined a meeting, or visited the deal room in 7+ days. Either the deal is in trouble or the champion has lost internal ground. Suppressed when CHAMPION_DEPARTED fires — that's a more specific signal.
BLOCKINGP5 · Sales Motion Maturity
Selected Vendor without Finance contact
Deal is at the budget-approval gate without a Finance/CFO contact on the OCR. Per case data, this is the stage where deals die.
BLOCKINGP4 · Stakeholder Engagement Sequencing
Selected Vendor without Procurement contact
Selected Vendor stage without a Procurement contact on the OCR. Contracting will stall when paperwork hits an unprepared procurement team.
BLOCKINGP4 · Stakeholder Engagement Sequencing
Evaluating without Finance contact
Strategic priority #4: surface bottom-of-funnel blockers at the top of the funnel. Get Finance involved BEFORE Selected Vendor, not after.
ACTIONP4 · Stakeholder Engagement Sequencing
Evaluating without IT/Security contact
Strategic priority #4. IT/Security reviews are a known multi-week unlock; getting them started during Evaluating de-risks Contracting timing.
ACTIONP4 · Stakeholder Engagement Sequencing
Demo Sat without outcome-first trial brief
Strategic priority #1 (the case's 'highest-leverage change available in H1'). Every Evaluating+ deal should have a trial brief delivered before the next meeting.
ACTIONP1 · Outcome-First Trial Motion
Finance contact exists but Finance brief not sent
Strategic priority #2 — the assets exist; the work is adoption. If a Finance contact is on the OCR but no Finance Meeting Brief has been delivered, we have a usage gap.
ACTIONP2 · Finance + IT Enablement Package
Security questions raised but IT one-pager not sent
Strategic priority #2. Champion has signaled IT scrutiny ('security team has questions') but the IT Zero-Lift one-pager hasn't been delivered.
ACTIONP2 · Finance + IT Enablement Package
Single-thread risk on Evaluating+ deal
Strategic priority #5 (sales motion maturity — multithreading is the focus area). Only one contact role attached to a meaningful-stage deal.
ACTIONP5 · Sales Motion Maturity
Deal aged past stage benchmark
Standard pipeline hygiene. Stage age > benchmark suggests either an unaddressed blocker or a forecast accuracy gap.
ACTIONP5 · Sales Motion Maturity
Qualified champion without demo booked
Champion identified at Qualified but no Demo Sat activity in the last 7 days. Conversion at this hop is 37% — slippage is expensive.
ACTIONP5 · Sales Motion Maturity
Last call surfaced risk markers
Most recent call transcript contains explicit risk markers (competitor mentioned, pricing pushback, no firm next step, etc.).
ACTIONP3 · AI-Powered Deal Execution Stack

P6 · ABM has no rules in v1 — on the rollout, not the architecture.

What we don't build in v1

  • Black-box deal-health scores (0–100). Named signals survive contact with reality; composite scores don't.
  • Auto-writes to the CRM. We earn the write before we take it.
  • Real-time everything. Morning digest is the right cadence for ~90% of signals. Real-time exists for the 10% that page.
  • Predictions. “67% chance to close” gets quoted in forecasts and held against the system when wrong. Stick to verifiable facts.

Rollout

Three phases. The first one decides if the rest happen.

Phase 1 · Weeks 1–3 — Win the wedge
3 signals. 1 surface. 1 pod.
SCOPE
  • 3 signals only, all mapped to Outcome-First Trial Motion.
  • 1 surface: daily Slack DM to the AE. No console, no manager view.
  • 1 pod: 3 AEs chosen for trust, not for need.
METRICS
  • Leading: % of late-stage deals with the missing stakeholder added — target +30 pts in 6 weeks.
  • Trust: signal mute rate < 10%, action rate > 50%.
  • Lagging (8-week look): conversion at the wedge stage vs control — target +10 pts.
Phase 2 · Weeks 4–8 — Earn the digest
Console + manager view + Studio v0.
SCOPE
  • Expand the rule library to the full v1 catalog.
  • Ship the unified console with the full morning digest.
  • Ship Manager view for one pilot pod.
  • Ship Studio v0 (RevOps-only).
  • Roll out to all AEs.
METRICS
  • Adoption: digest open rate > 80% within 3 days for each new AE.
  • Adoption: action-taken-within-72-hours on top-3 signals > 50%.
  • Studio: ≥ 5 rules authored via Studio in 4 weeks.
  • Lagging: AE time-in-CRM down 20%, time-in-customer-conversations up.
Phase 3 · Weeks 9–12 — Compound the loop
Cross-deal patterns + ABM signals + feedback flywheel.
SCOPE
  • Open Studio to managers — codify questions instead of repeating them.
  • Cross-deal pattern signals: champion-changing-jobs, competitor mention clusters.
  • Account-level ABM signals — intent scoring, multi-deal aggregation.
  • Feedback dashboard — low action-rate rules auto-flag for review.
  • Begin selective write-back scoping.
METRICS
  • Rule velocity: 2+ shipped per week, < 1 deprecated per week.
  • Compounding adoption: action-rate increases over time.
  • Lagging: win-rate uplift on Eval+ deals where ≥ 2 recommended actions were taken.

The single metric I'd defend in the boardroom

% of late-stage deals with the missing stakeholder role added within 7 days of the signal firing.

Cleanest proxy for whether the system is changing behavior at the workspace's named kill point (Deals die at Selected Vendor: budget approval fails because Finance/IT were engaged too late.). Instrumented in the CRM already. Leading indicator for the lagging metric that matters: conversion through the kill zone.

Companion system: Trial Orchestrator

The signal engine surfaces “deploy a trial.” This companion removes every step between that decision and the trial being live in 48 hours.

The system in 6 steps

01
Trigger

AE clicks "Start Trial" on an Eval opp. Auto-fires when the signal engine raises a no-trial-brief warning.

02
Intake (5 min, AE)

Auto-prefilled form pulls account, champion, recent call summaries, deal-room engagement. AE confirms 3 KPIs.

03
SE assignment + clock

Round-robin by segment + load. 48-hour SLA, visible to AE, SE, and manager.

04
KPI Assessment + Pre-Seeded Demo (SE, ≤ 48h)

SE builds the assessment, seeds the demo with the buyer's scenarios, drops both into a Dock room.

05
Champion enablement package (auto)

Dock auto-populates: KPI Assessment, Pre-Seeded Demo, Finance Meeting Brief, IT Zero-Lift One-Pager, similar customer story, ~90s AI-generated walkthrough video.

06
Champion notification (auto)

Email + Slack to champion. Engagement tracked back to the deal in Salesforce.

Why it compounds with the rest of the intelligence layer

  • Signal detects the gap. Orchestrator removes the work between detection and delivery. Together: “we should run a trial” → “trial is live” in two days.
  • Operationalizes existing assets (Outcome-First Trial Brief, Finance Meeting Brief, IT Zero-Lift One-Pager) at the moment they matter.
  • Cheap to ship: CRM flow + cron jobs + deal-room API + one prompt. ~3 sprint weeks.

Measurement (4-week pilot)

Trial-deploy lead time
~6 business days → ≤ 48 business hours on 80% of triggered trials
Trial deployment rate
~35% of Eval deals → 90% in 8 weeks
Trial → won conversion
Hold steady at 3× the trial volume — proves quality isn't degrading
SE utilization
Past 80%, that's the data to justify SE #3