Operational Framework

20 steps, 8 AI agents.
Every phase in detail.

The 4 stages (Diagnose, Build, Deliver & Adopt, Iterate) are powered by a structured operational framework. Here's every phase — from onboarding to iteration.

01

Onboarding

Steps 1–4

Roles, communication channels, and 7 meeting templates — before any work begins.

Concept, roles, communication, meeting formats

Every person gets a defined role matched to their profile and strengths. 7 meeting templates set — each with a fixed agenda, escalation mechanism, and clear output.

RolesProblem Solver · Task Implementer · Domain Problem Owner · Interim Lead
TemplatesOperational (2×/week) · Weekly strategic · Kickoff · Acceptance · Adoption checkpoint · Team assembly · Retro
02

Planning

Steps 5–7

Meeting cadence locked in. One priority selected. No exceptions.

Meeting cadence & priority selection

Operational meetings locked in (Tue/Thu 11:00). Weekly strategic reviews scheduled. Leadership selects one priority — not two "because both are urgent."

The Problem Solver operationalizes it — turns a business need into a defined scope with success criteria.

RuleOne priority per team at a time. No exceptions. (Codex, principle 17)
03

Mobilization

Steps 8–10 3 AI

Dedicated team assembled. Structured kickoff. Domain knowledge captured by AI.

Team assembly

Step 8

Dedicated team: Problem Solver (diagnoses what & why), Task Implementer (builds the solution), Domain Problem Owner (owns context, accepts results).

Domain Problem Owner gets a 10-minute onboarding: how we collaborate, communicate, what we expect, how much time it takes.

Kickoff

Step 9
AI Kickoff Manager AI Meeting Analyst

AI generates meeting structure, task breakdown per role, MVP definition, first-week goals. Action items extracted within 15 minutes.

Domain knowledge capture

Step 10
AI Domain Briefer

AI structures every insight into a reusable knowledge base: domain summary, actor map, process descriptions, business rules, data sources, exceptions, gaps, glossary.

Nothing falls through the cracks. This knowledge base is reused in every subsequent iteration.

04

Execution

Steps 11 2 AI

Iterative build. Disposable technology. Smart questions, not surveys.

Build — iteratively, in parallel with daily operations

Step 11
AI Meeting Analyst AI Problem Explainer

Task Implementer starts with an MVP — fast, real data, disposable technology. Value is in solving the business problem, not in code elegance.

When input is needed, AI generates focused questions with options A/B/C. The Domain Problem Owner picks — doesn't write an essay. Blockers escalated immediately.

PhilosophyBuild for replacement, not for eternity. (Codex, principle 1)
EscalationTechnical blocker → immediate. PS-TI disagreement → Interim Lead. No acceptance after 4 weeks → scope review.
05

Acceptance

Steps 12 2 AI

Formal sign-off. Change Card generated — organizational memory of every change.

Domain Problem Owner says "this solves my problem"

Step 12
AI Change Card Writer AI Problem Explainer

Explicit confirmation — a real checkpoint, not a nod in passing. Rejected 3+ times → priority escalated.

Upon acceptance, AI generates a Change Card — the organizational memory of every delivered change.

Where it's connectedHow it's maintainedHow to report issuesBusiness context gainedLinks & referencesTest data
RuleNo Change Card = no formal acceptance. (Codex, principle 3)
06

Adoption

Steps 13 1 AI

Delivering code is 50%. Adoption is the other 50%. Measured at 1 and 2 weeks.

Adoption support & measurement

Step 13
AI Adoption Writer

AI generates tailored adoption materials. Problem Solver runs two mandatory checkpoints.

1-weekIs the team using it? What's blocking them?
2-weekFull assessment. Usage, active users, time reduction. Below target → escalated.
MetricWe measure adoption, not deployment. (Codex, principle 14)
07

Reporting

Steps 14 1 AI

Board-ready Executive Summary. Maintenance model established.

Executive Summary

Step 14
AI ExSum Writer

AI aggregates Change Card, adoption metrics, and strategic context. Generated after the 2-week checkpoint.

Maintenance model established: first technical contact, issue reporting, quick fix (<1h) vs. iteration loop.

08

Iteration

Steps 15–17 1 AI

Change request → AI diagnosis → replan → back to execution.

Change request → Diagnosis → Replan → Build

Steps 15–17
AI New Feature Briefer

Domain Problem Owner reports a change. AI aggregates Change Card + adoption data + domain knowledge to diagnose.

IterationSame owner, same domain → extend what exists. Back to step 11.
New priorityDifferent scope → back to step 7.
RuleIterate on what exists, don't rebuild from scratch. (Codex, principle 4)
09

Cross-cutting

Steps 18–19

Continuous activities that run throughout every engagement.

Employee fear reduction

Step 18

Every deployment triggers an assessment. Proactive communication, training, transparency. Show what changes, what doesn't.

AI development roadmap

Step 19

After every Executive Summary, evaluate whether AI agent responsibilities should expand.

10

Reflection

Steps 20

Retro — 30 min, no slides. Both teams. Then: next priority.

Retro

Step 20

Team shares what worked and what didn't. Other team listens. Process insights fed back into the framework.

Next priority from backlog → back to step 7