Ship in phases.
Useful on day one.
Six phases. Each self-contained. Each ends with the team actually using what was built — not a demo, not a "soft launch," a genuine handoff. Nobody waits twelve months to see value. We can stop after any phase and the work done already stands on its own.
Big-bang AI projects kill themselves.
The default shape of an AI transformation engagement is: six months of integration work, three months of model work, one big cutover, followed by a month of on-the-ground firefighting while operators learn the new system. That shape fails, consistently, because it puts every risk in the last mile.
The framework inverts this. We sequence so the least-risky, most-visible wins ship first — data model, pipeline stages, intake migration — and the AI work comes after the ground is clean. If the program gets defunded in month four, the client still has a working pipeline. Every phase ends with a hard usable-after milestone we'll sign our names to.
Every phase has a written "usable after" line in the statement of work. If what shipped doesn't deliver that outcome, we fix it before starting the next phase. No exceptions.
Sequenced for value, not for convenience.
Indicative hour totals are for a pipeline-heavy services firm with one active pipeline and a roster of 5–10 operators. Scales linearly with additional pipelines.
Foundation & board
The data model, the stages, and the card view. The least glamorous phase; the phase everything else builds on.
- 21 custom properties on the Deal and Contact objects
- Pipeline definition + 8 deal stages with stage gates and rotting-timer configuration
- Deal-card view (doc progress bar, conditions, owner, gate status) + Contact-record layout
- 4 core email templates (Step 1, Expectations, Final Warning, Timeline)
- Board walkthrough with operators — create test deals, verify layouts
Intake + deal creation
Kill the spreadsheet. Wire the log-meeting trigger. Operators never open Drive again.
- Field audit, CRM Contact properties, CRM-native intake form on the Contact record
- Legacy Drive template deprecated and archived
- W0b enhanced deal-creation workflow — trigger on log-meeting, auto-populate deal fields, Step 1 + 24h Expectations emails
- Google Chat webhook for deal-created alert
- End-to-end testing + operator training on the new intake flow
Pre-submission workflows
Strategy review, contract delivery, document collection, rotting alerts, timeout cleanup.
- W1 Strategy Session entry + rotting (Round 2+ re-entries only)
- W2 Contract delivery on stage move
- W3 Document-collection email sequence (4-week cadence with re-enrollment — the most complex workflow in this phase)
- W4 Rotting-timer alerts across all stages (5d / 30d / 2d / 15d / 10d)
- W5 30-day document timeout + lead-status flip
- 3 document-reminder templates + Google Chat webhook wiring + UAT
AI document processor
The 11-step AI pipeline that owns the "client signed documents" stage end-to-end.
- E-signing platform API integration (daily polling + webhook) + CRM API tooling for the agent
- Source-document PDF extraction — conditions, service-connection status, dates
- Condition reconciliation + mismatch detection with Contact-note updates
- Form-population engine — PDF form-field mapping across the three forms
- Client-email drafting + LangGraph orchestration + dead-letter queue
- UAT with 5+ real test clients + operator training on AI-form review
Post-submission + AI router
The full case lifecycle. Post-submission cadence plus the AI router that reads inbound notifications and moves the deal.
- W6 Submitted cadence (Timeline, exam scheduling, Day 40, Day 70)
- W7 Close-date automation · W8 Round X re-entry setup
- 3 post-submission email templates
- W10 integrations: Trello webhook + shared-folder watcher
- Notification PDF parsing — classifier for 4 notification types
- 4 notification branches: Dependent · Exam · Favorable (invoice) · Adverse (escalation)
- Routing logic (Won / Lost / Deferred / no-change) + idempotency + UAT
Reporting & dashboards
Seventeen KPIs across pre- and post-submission, role-based dashboards for every persona.
- 11 pre-submission KPI reports (meetings booked, deals created, deal value, time per stage, lead source…)
- 6 post-submission KPI reports (time in Pending, Won / Lost / Deferred counts, submission-to-close duration)
- 4 role-based dashboards — Operator, Manager, Finance, Senior Operator
- Event-tracking hooks at critical stage transitions
- Report UAT + management dashboard walkthrough
Four rules we do not break.
These are contract-level commitments. They go in the SOW, not in a slide deck.
Usable-after is signed
Every phase has a written usable-after milestone in the SOW. If we don't deliver that outcome, we fix it before the next phase starts. No "we'll come back to it."
UAT ends every phase
Hands-on UAT with real operators and real cases, not a demo. The phase is "complete" when the operators use it, not when we say it's done.
We ship data before AI
The AI-owned phases are always 3+. We don't put an agent on top of a messy data model, ever. Garbage in, expensive garbage out.
Stop is a valid outcome
After every phase, the client can stop the program. The work already done stands. We've never held a final phase hostage to force a renewal.
17 KPIs shipped with the build.
Management dashboards are not an afterthought — they're Phase 6. Every KPI below is live, role-based, and tied to the same data the operators see on their board.
- 01Meetings booked by operator, by day
- 02Meetings rescheduled / no-show by operator
- 03Deals created by day / month, by operator
- 04Round 1 first-time deals
- 05Round 2+ supplemental deals
- 06Deal value by operator (R1 vs. R2+ mix)
- 07Time spent per stage by operator (bottleneck view)
- 08Time from session to signed contract
- 09Customer interactions per deal, per operator
- 10Contacts by lead source (paid, organic, referral)
- 11Lead-to-client conversion by channel
- 01Time in Pending Submission by operator
- 02Submissions per operator, per day / month
- 03Supplemental submissions per operator
- 04Time spent per stage, stages 4–8
- 05Deals Won / Lost / Deferred per operator
- 06Submission-date to close-date duration by operator
We build on the stack we hold partner status in.
AWS Select Consulting Partner · Anthropic-aligned for Claude · HubSpot Solutions Partner · Salesforce · Google Cloud · Microsoft Azure · Shopify · ISTQB.
Six phases. Useful from phase one.
This is how we run every pipeline-automation + AI engagement. Written into the SOW, signed on both sides, shipped phase by phase. If we scope it, you get the usable-after.