Implementation · Ch. 16

Your first value stream
in 90 days.

Ninety days will not make you Level 4. It will put one governed, end-to-end AI capability into production: the proof point that the architecture works. From there, Level 3 by the end of year one and Level 4 by the end of year two. This is the canonical phase-by-phase, week-by-week roadmap, and the discipline of the first thirty days is what decides the rest.

Phase 01 · D01–D30 Foundation: team & decisions
Phase 02 · D31–D60 Build: Data Fabric + Governance
Phase 03 · D61–D90 Demonstrate: first value stream
D30 ·
Charter, executive coalition, architecture decisions, honest baseline. Build begins Week 5.
D60 ·
Data Fabric and Governance live in parallel; platform stood up.
D90 ·
One production-ready capability, end to end. The architecture is proven.
01 · Precondition Genuine CEO sponsorship and a coalition of CFO, CRO, and CHRO, before Day 1.
02 · Principle Governance is built in parallel with the Data Fabric, never bolted on afterward.
03 · Principle The first thirty days are organizational. Get the decisions right or pay for them in month four.
Phase 01 · Days 1–30Foundation

The first thirty days
are organizational.

No architectural build yet. The deliverables are the team, the governance structure, the vendor relationships, the architectural decisions, and an honest baseline. The architectural build begins in Week 5. Rush past Day 30 and the next sixty days produce work that has to be redone.

Week 01 · Launch

Executive launch, vendors, roles

  • CEO convenes the five-person coalition; signed program charter distributed to the expanded leadership group.
  • Vendor scouting: shortlists for cloud AI services, data platform, streaming backbone, MLOps/LLMOps.
  • Role identification: Head of Data Fabric, Head of AI Platform, Head of Governance Engineering, and the first Lead AI Product Managers.
Week 02 · Coalition

Coalition & governance cadence

  • Individual commitments from the COO, CMO, General Counsel, and the business-unit leaders who will host the first capabilities.
  • First Steering Committee: cadence, decision rights, escalation paths agreed.
  • Honest maturity baseline confirmed (external assessment if the internal one isn't candid).
Week 03–04 · Decisions

Architectural decisions, locked

  • Streaming backbone, data catalog, MLOps/LLMOps platform, foundation-model strategy, agent orchestration — recorded in an architectural decision record the Steering Committee approves.
  • Decisions are not revisited in the first 90 days except on vendor failure. Stability is what lets the build move at speed.
  • Offer letters signed for the critical roles.
Day 30 milestoneD30 ✓
Charter, executive coalition, Steering Committee operating model, architectural decisions, role roster, vendor shortlists, and an honest baseline — all in place. The architectural build begins in Week 5.
Phase 02 · Days 31–60Build

Data Fabric and Governance,
built together.

Layer 1 and Layer 4 run in parallel from Week 5 — governance cannot be retrofitted onto the data fabric, and the fabric cannot operate at scale without it. Layer 2 begins in Week 7. By Day 60 the architectural foundation is operational.

Week 05–06 · Layer 1

Data Fabric: first three domains

  • Streaming pipelines for the first three data domains the priority use cases require.
  • The semantic layer (canonical entity definitions) and the first version of the catalog.
  • Data-quality monitoring (Monte Carlo / Great Expectations) running against all three.
Week 05–06 · Layer 4 (parallel)

Governance: policy library & AI inventory

  • First version of the policy library, drawing on NIST AI RMF, EU AI Act, and sector frameworks (FDA, SR 11-7).
  • Complete AI inventory with risk classification for every current capability; high-risk ones flagged.
Week 07–08 · Expand + Platform

Data products, controls, and Layer 2

  • Layer 1 expands to five domains and ships the first three data products with documented SLAs.
  • Layer 4 implements controls for the high-risk capabilities; audit evidence begins generating.
  • Layer 2 begins: foundation-model strategy, the model gateway, the MLOps/LLMOps platform stood up end to end.
Day 60 milestoneD60 ✓
Layer 1 with five data domains and three data products; Layer 4 with policy library, AI inventory, and controls operating; Layer 2 with foundation-model access and the MLOps platform stood up. The foundation is operational.
Phase 03 · Days 61–90Demonstrate

The first value stream,
end to end.

A Layer 3 capability that draws on Layer 1 data, runs on Layer 2, operates under Layer 4 governance, and produces value Layer 5 can attribute. Day 90 is the proof point, not the finish line: the full architecture takes two to three years to mature.

Week 09–10 · L03 + L04 + L05

First capability, governance review, value method

  • The first AI Product Triad (PM + AI Engineer + Domain UX) ships a clean-attribution capability: fraud detection, supplier risk, customer-service routing, contract analysis, or expense audit — in a bounded pilot scope, with trust UI and human-in-the-loop controls.
  • Layer 4 runs the first quarterly governance review across the portfolio.
  • Layer 5: the AI Value Office drafts the AI Contribution Scorecard methodology the CFO can defend.
Week 11–12 · Production-ready

Harden and measure

  • The capability moves to production-ready: operational metrics validated, user feedback incorporated, workflow integration tested at scale.
  • Initial value attribution captured (cycle time, cost per unit, quality); disclosure workflow and audit trail operating.
Day 90 · Readout

The proof point, on the record

  • One-page and fifteen-page status reports to the executive sponsors; a twenty-minute readout to the Board.
  • Decision records preserved as the program's institutional memory. The next ninety days are scoped.
Day 90 milestoneD90 ✓
Five data domains and three data products; foundation-model access and MLOps platform; one production-ready AI capability; policy library, AI inventory, and a completed governance review; and an approved value-attribution methodology. The architecture is operating end to end.
Beyond the quarter

Months 4–24:
scale, standardize, autonomize.

Day 180 Roll the proven playbook Layer 1 extends to fifteen domains and ten data products; the second through fifth capabilities deploy, with the second in full production. Second Quarterly Value Review documented.
Year one · Day 365 Reach Level 3 The architecture operates across all five layers. The AI Center of Excellence is in place. The CFO can defend quarterly value figures. The gap to the laggards begins to widen.
Year two · Day 730 Reach Level 4, then push to 5 Value at the scale that justifies the investment; the industry's leaders recognize you as a peer. From there, self-optimizing agents and the compounding Level-5 advantage.
Begin

Run the assessment first.
Then start the clock.

You cannot plan a 90-day roadmap without knowing where you stand. The assessment is 10 minutes. The plan is yours.

Run the assessment → See the framework →