Your 90-Day Action Plan

Chapter 5 — Your 90-Day Action Plan

_ □ ×

Your 90-Day Action Plan

Three eras. Twenty-three milestones. One quarter to transform your team.

Every advance in the tech tree below maps to a concrete action you can take this week. Research them in order or jump around. Click a node to mark it complete, unlock the next era by finishing six in the current one. When you're ready to make the case, grab the copyable plan at the bottom and send it to your CTO.

🏛 civilize.exe — Technology Advisor

_ □ ×

0 of 23 advances discovered

Phase 1: Foundation (Days 1-30)

Audit test coverage and identify critical paths below 80%
Set up a project instruction file (called CLAUDE.md) that teaches the AI agent your codebase's rules and conventions
Choose one senior developer as the AI pilot
Install and configure agentic tools (Claude Code, Cursor, etc.)
Pre-approve routine operations (running tests, checking code style) so the agent doesn't stop and ask permission for every small step
Pick one well-defined feature for the first AI-assisted build
Measure baseline: how long does a typical feature take today?
Establish AI code review guidelines (what to look for in AI output)

Phase 2: Expansion (Days 31-60)

Roll out agentic tools to the full development team
Train team on effective prompting and agent oversight
Integrate AI code review into your PR process
Set up automated security scanning for AI-generated code
Set up automatic quality gates that check the agent's code for errors and style violations on every edit
Start tracking AI-assisted vs. manual feature completion times
Identify codebase gaps that slow down AI (missing tests, types, docs)
Begin filling those gaps as part of regular sprint work

Phase 3: Optimization (Days 61-90)

Train senior devs on running parallel AI sessions (3-5 concurrent)
Define project-specific AI conventions (what patterns to follow, what to avoid)
Set up automated AI-assisted security reviews
Redesign sprint planning to account for AI-boosted velocity
Document your team's AI playbook (what works, what doesn't)
Calculate ROI: effective output vs. tool costs vs. baseline
Plan next quarter's roadmap based on new velocity

90-DAY AGENTIC CODING ROLLOUT
=============================

PHASE 1: FOUNDATION (Days 1-30)
Goal: Prove it works with one engineer.

[ ] Audit test coverage, find critical paths below 80%
[ ] Set up CLAUDE.md (project instruction file for the AI agent)
[ ] Choose one senior dev as the AI pilot
[ ] Install agentic tools (Claude Code, Cursor, etc.)
[ ] Pre-approve routine ops (tests, linting) so the agent runs uninterrupted
[ ] Pick one real feature for the first AI-assisted build
[ ] Measure baseline: how long does a typical feature take today?
[ ] Establish AI code review guidelines

Success: Pilot completes one feature with AI. You have baseline numbers.

PHASE 2: EXPANSION (Days 31-60)
Goal: Full team adoption, quality gates in place.

[ ] Roll out agentic tools to the full dev team
[ ] Train team on effective prompting and agent oversight
[ ] Integrate AI code review into PR process
[ ] Set up automated security scanning for AI-generated code
[ ] Add quality gates (lint, type-check, test on every AI edit)
[ ] Start tracking AI-assisted vs. manual completion times
[ ] Identify codebase gaps that slow the AI (missing tests, types, docs)
[ ] Begin filling those gaps in regular sprint work

Success: Full team using AI daily. Feature time down 30-50% vs. baseline.

PHASE 3: OPTIMIZATION (Days 61-90)
Goal: Maximize throughput, quantify ROI.

[ ] Train senior devs on parallel AI sessions (3-5 concurrent)
[ ] Define project-specific AI conventions (codified in writing)
[ ] Set up AI-assisted security reviews (deeper integration)
[ ] Redesign sprint planning for AI-boosted velocity
[ ] Document the team's AI playbook (what works, what doesn't)
[ ] Calculate ROI: effective output vs. tool costs vs. baseline
[ ] Plan next quarter's roadmap based on new velocity

Success: Two-sprint features ship in one. ROI quantified for leadership.

---
Source: gabezen.com/guide — The CTO's Guide to Agentic Coding

## Sprint Planning — AI-Augmented Workflow

### Story Classification
Tag each story before estimation:

**AI-SOLO** - Agent can complete with minimal oversight
  Examples: CRUD endpoints, unit tests, data migrations,
  boilerplate components, documentation updates
  Estimate: 10-20% of pre-AI estimate

**AI-ASSIST** — Developer leads, agent accelerates
  Examples: New features, refactoring, integrations,
  bug fixes with clear reproduction steps
  Estimate: 30-50% of pre-AI estimate
  Note: Open-ended or greenfield work fits here, but
  requires detailed PRDs/specs upfront. The harder the
  problem, the more the effort shifts from coding to
  specifying. Budget time for spec writing accordingly.

**HUMAN-ONLY** — Requires human judgment throughout
  Examples: Architecture decisions, security-critical code,
  performance optimization, vendor evaluations
  Estimate: 80-100% of pre-AI estimate

### Capacity Planning Adjustments
- Senior dev with AI: 2-3x previous velocity on AI-SOLO/ASSIST
- Junior dev with AI: 1.2-1.5x (still learning to review)
- AI-SOLO stories can run in parallel (batch 3-5 per dev)
- Budget 20% of sprint for AI infrastructure improvements

### Sprint Review Additions
- Track AI-assisted vs. manual completion times
- Review AI-generated code quality (bugs found post-merge)
- Update CLAUDE.md with new patterns discovered
  (e.g., agent keeps starting the dev server when it's
  already running, add one line to CLAUDE.md and it
  never happens again. Encode the fix, not the complaint.)
- Share effective prompts across the team

Most teams can research Era I on their own. Where it gets tricky is the codebase-specific stuff: which tests to write first, how to structure the CLAUDE.md for your architecture, what conventions the agent needs to be productive. That's where having someone who's done it before saves you weeks of false starts.