Skip to module content
🧠AI Academy
0/91
Sign in
Module 22 · ~6 min

The 90-Day Implementation Roadmap

Days 1–30 Foundation · 31–60 Systems · 61–90 Market.

Reading progress
0/8 · 0%

The big idea

💡Key idea
The 90 days break into three phases that must be run in order. **Days 1–30 Foundation:** consolidate MCP behind a gateway with per-client profiles and audit logging, parameterize the Google Ads suite, write the first six internal skills, stand up a prompt/skill registry and minimum eval harness, install and fork the foundational repos. **Days 31–60 Systems:** build the deep-research skill and subagent orchestration, port the LinkedIn project to a portable skill chain, assemble Sprints 1 and 3 as productized packages, apply the full GEO playbook to your own site, and ship the demo bench with recorded fallbacks. **Days 61–90 Market:** publish a flagship original-research asset, run Sprint 0 and Sprint 1 with two existing clients as documented case studies, package the Digital Twin offer, launch the OS Operations retainer to past clients, and run the first quarterly stack audit. Sequencing matters. Publishing without a demo bench wastes inbound. Building systems without a foundation wastes effort.
Quick check
1 question · instant feedback
0/1
  1. The 90-day sequence is:

Numbers that matter

6
Internal skills written in the first 30 days: sprint-loop, client-onboarding, weekly-client-report, brand-voice, substantive-pushback, pre-meeting-brief.
2
Documented case studies (Sprint 0 + Sprint 1 on existing clients) by Day 90.
Quick check
1 question · instant feedback
0/1
  1. The first 6 internal skills to write are:

Deep dive

3/3 open

The Foundation phase is about infrastructure and internal tooling — nothing client-facing yet.

Consolidate your MCP stack behind a gateway with per-client profiles and audit logging. Port the Google Ads suite to parameterized per-client form and mirror the architecture for Meta.

Write the first six internal skills: sprint-loop, client-onboarding, weekly-client-report, brand-voice, substantive-pushback, pre-meeting-brief. These six accelerate every future engagement.

Stand up a prompt/skill registry in git with a minimum eval harness. These are the governance rails that keep the library from becoming chaos.

For repos: install, vet, and fork anthropics/skills, superpowers, agents-best-practices, and Ay-Skills. Fork everything you adopt — vendor and pin, so you're never surprised by upstream changes.

The Systems phase builds the capabilities that go into sprint packages.

Build the deep-research skill with subagent orchestration; wire Firecrawl, Exa, and Tavily as the retrieval layer.

Port the LinkedIn Project into a portable skill chain: hook library, atomizer, learning loop — the Content OS core.

Assemble Sprint 1 and Sprint 3 as productized packages: scope docs, runbook templates, demo sandboxes ready to show.

Apply the full GEO playbook from Module 11 to your own site. Start the weekly citation tracker. You are your own case study — this can't wait.

Ship the demo bench from Module 21 with recorded fallbacks. Systems that exist but can't be demoed don't close deals.

The Market phase turns built systems into pipeline.

Publish a flagship original-research asset — something like 'The State of AI Operating Systems in B2B Marketing: data from N workflow audits.' This single asset does three jobs simultaneously: GEO citation magnet (structured, extractable, citable), LinkedIn pillar content, and sales proof for the sprint catalog.

Run Sprint 0 and Sprint 1 with two existing clients and document the results with hard numbers. These become your case studies for everything that follows.

Package the Digital Twin offer from the DTI build, anonymized, as a retainer flagship. Launch the OS Operations retainer to every past client.

End the quarter with the first stack audit. Kill what didn't earn its place. Sprawl is the failure mode that compounds in the wrong direction.

Quick check
1 question · instant feedback
0/1
  1. The Day 90 flagship research asset serves:

How to run it

  1. Foundation before Systems before Market
    Sequencing matters.
  2. Fork every adopted repo
    Vendor + pin, day 1.
  3. GEO on your own site by Day 60
    You are your own case study.
  4. Quarterly stack audit is mandatory
    Kill what didn't earn its place.
Quick check
1 question · instant feedback
0/1
  1. GEO on your own site by Day 60 is:

In the field

🔬Worked example
Days 61–90 Market — flagship original-research asset ('The State of AI Operating Systems in B2B Marketing — data from N workflow audits'): simultaneously a GEO citation magnet, LinkedIn pillar, and sales proof. One asset serving three funnels.
🚫When not to reach for it
Don't sequence market work (Days 61–90) before Systems (Days 31–60) are shipped. Publishing the flagship research without a demo bench to convert inbound leads is wasted attention.
Quick check
1 question · instant feedback
0/1
  1. The quarterly stack audit exists to:

Pitfalls & takeaways

Failure modes

  • Running quarters like calendars — no compounding.
  • No first-quarter stack audit — kills what didn't earn its place is skipped.
  • Original research is anecdote-driven, not audit-data-driven — no defensibility.

Durable takeaways

  • Sequence matters: Foundation → Systems → Market.
  • GEO on your own site is non-negotiable — you are your own case study.
  • The quarterly stack audit is what keeps the stack lean.
  • One flagship research asset can serve three funnels at once.

Do the work

🏋️Prove you learned it

Write your own 90-day plan against this template. Ship one Days 1–30 milestone this week (e.g., stand up the gateway with 2 client profiles, or write the sprint-loop skill). Document what you'll kill in the quarterly audit.

0 chars
📦Artifact to produce
90-day plan with weekly milestones + a first-quarter stack-audit template.

Sources

  • · GMS Field Manual §22 (90-Day Roadmap)