AI Transformation Roadmap

Use this when an organization needs a sequenced, executive-ready plan for adopting AI across the business, not a single use case and not a strategy deck. It takes current-state maturity, a scored portfolio of candidate use cases, and the capability and governance gaps, then sequences them into Now, Next, and Later phases where every phase names its path to production. If you only need to assess where an org stands today, use /ai-maturity-org first and feed its output in here.

Related skills: Grounds current state with /ai-maturity-org and pressure-tests readiness with /ai-adoption-evaluator. Borrows the Now/Next/Later structure from /now-next-later-roadmap. Pairs with /fractional-caio-playbook for the operating model that carries the roadmap.

The hard part most teams miss

In 2026 the strategy deck is the easy part. Enterprises are past generative pilots and are moving to embedded, agentic AI tied to business outcomes. The demand now is execution and governance, not vision. A roadmap that sequences use cases and stops there will stall.

1. Roadmaps die in pilot purgatory. Most organizations run many proofs of concept and ship almost nothing. A demo that wows a steering committee is not a system real users depend on. Every phase in this roadmap must name a concrete path to production: who owns it, what integration and security review it clears, what "in production" means as a checkable condition. A phase whose only deliverable is a pilot is a phase that will not move.
2. Capability and governance ride alongside use cases, or every project re-solves the same plumbing. If platform investment waits until "after the pilots prove value," each team rebuilds evaluation harnesses, data pipelines, access controls, and monitoring from scratch. The second use case should be cheaper than the first because the first one built reusable capability. Sequence the platform deliberately, in parallel with the use cases that fund it.
3. The bottleneck is adoption and governance, not the model. The models are good enough. What stalls transformation is people not changing how they work, leaders not trusting outputs they cannot audit, and no operating model for who approves, monitors, and owns AI in production. A roadmap that ignores change management and the governance build is a roadmap that ships capability nobody uses.

Process

Step 1: Gather inputs

Ask the user:

1. Organization and scope. {{org_name}} and which business units or functions are in scope for this roadmap.
2. Current state. Maturity level, data readiness, existing AI capability, and the biggest governance gaps. If an /ai-maturity-org assessment exists, use it; if not, capture a quick read per dimension.
3. Candidate use cases. {{candidate_use_cases}}, usually 8 to 20 ideas across functions. Rough description and the outcome each would move.
4. Strategic outcomes. {{target_outcomes}}, the 2 to 4 measurable business results the transformation must drive (revenue, cost, cycle time, risk).
5. Constraints. Budget envelope, regulatory or compliance requirements, data residency, existing platform commitments, and realistic capacity.
6. Time horizon. How far out the roadmap should reach (typically 12 to 18 months across three phases).
7. Risk appetite. How tolerant the org is of customer-facing AI, autonomous action, and being early.

If input is partial, proceed with what is available and list assumptions explicitly.

Step 2: Read the current state honestly

Summarize where the org actually is across four dimensions. Be specific; vague maturity ratings produce vague roadmaps.

• Maturity: experimenting, scaling, or operating. Most orgs overstate this.
• Data readiness: is the data the use cases need accessible, governed, and trustworthy, or does the roadmap need a data workstream before anything ships.
• Capability: what platform, talent, and tooling exist today versus what the portfolio requires.
• Governance: who currently approves, monitors, and owns AI. If the answer is "nobody," that gap goes into Phase 1, not Phase 3.

Name the single biggest constraint that will limit pace. The roadmap sequences around it.

Step 3: Score the opportunity portfolio

For each candidate use case, score two axes and capture the path-to-production reality:

• Value: contribution to the target outcomes, High / Medium / Low. Tie to a number where one exists.
• Feasibility: data, capability, and governance readiness today, High / Medium / Low. A high-value use case the org cannot yet build safely is not a Now item.
• Path to production: what it takes to run this for real, not as a demo. Integration points, the review it must clear, the owner.
• Reusable capability it creates: what platform or governance asset this use case builds that later use cases inherit.

Plot value against feasibility. High value plus high feasibility are quick-win candidates. High value plus low feasibility are Later bets that the platform and governance work must unlock first. Low value drops off the roadmap; say so plainly.

Step 4: Sequence to avoid pilot purgatory

Sequence in three moves, not as a feature wishlist:

• Now (quick wins that reach production): a small number of high-feasibility, high-value use cases that can ship to real users this phase. Each must clear governance and integration, not just demo well. These earn credibility and budget for what follows.
• Next (reusable platform and capability): the shared assets that make every later use case cheaper, evaluation harnesses, data pipelines, access and audit controls, monitoring, the governance operating model. Fund this with the credibility the Now phase earned. A second wave of use cases rides on top of it.
• Later (scale and higher-risk bets): the high-value, lower-feasibility use cases the platform and governance now make viable, plus broader rollout. This is where agentic and customer-facing work belongs once the foundation holds.

Throughout, capability and governance build in parallel with use cases. Never schedule the entire platform "after" the use cases or "before" them; thread it through every phase.

Step 5: Generate the phased roadmap

Produce the roadmap in this structure. Each phase carries its use cases, the capability and governance it builds, and how success is measured.

# AI Transformation Roadmap: {{org_name}}

**Scope:** (business units / functions)
**Strategic outcomes:** (2-4 measurable results)
**Time horizon:** (e.g., 12-18 months)
**Last updated:** (date)

## Current state
- Maturity: (experimenting / scaling / operating)
- Data readiness: (summary + biggest gap)
- Capability: (what exists vs. what's needed)
- Governance: (who owns / approves / monitors today)
- Pace-limiting constraint: (the one thing that sets the speed)

## Phase 1 -- Now: quick wins that reach production
**Theme:** prove value with use cases real users depend on.
| Use case | Outcome link | Value x Feasibility | Path to production | Owner |
|---|---|---|---|---|
| (use case) | (outcome) | High / High | (integration + review + "done" condition) | (name/role) |

- **Capability built this phase:** (reusable assets, even small ones)
- **Governance built this phase:** (approval path, monitoring, ownership for what ships)
- **Change management:** (who adopts this, how they're brought along)
- **Measurement:** (the metric that proves this phase worked, with a baseline and target)

## Phase 2 -- Next: reusable platform and capability
**Theme:** build the foundation so the next use cases cost less than the first.
| Investment | What it enables | Use cases riding on it | Measurement |
|---|---|---|---|
| (platform/governance asset) | (capability unlocked) | (downstream use cases) | (readiness or adoption signal) |

- **Use cases shipping this phase:** (second wave, now cheaper to build)
- **Governance maturing this phase:** (operating model, audit, risk controls)
- **Change management:** (scaling adoption beyond early teams)
- **Measurement:** (platform reuse, time-to-production for new use cases, outcome movement)

## Phase 3 -- Later: scale and higher-risk bets
**Theme:** unlock the high-value work the foundation now makes viable.
| Use case / initiative | Outcome link | Why now viable | Trigger to start |
|---|---|---|---|
| (use case) | (outcome) | (which platform/governance asset unlocked it) | (event or metric) |

- **Capability at scale:** (what running broadly requires)
- **Governance at scale:** (audit, compliance, autonomous-action controls)
- **Change management:** (org-wide operating model, role changes)
- **Measurement:** (the transformation-level outcome, tied to the strategic results)

## ROI and measurement model
- (Outcome 1) -- baseline (X) -> target (Y), measured by (signal), reviewed (cadence)
- (Outcome 2) -- baseline -> target, measured by, reviewed
- Investment: (rough cost by phase) against (expected return)

## Risks and watchpoints
- (Top adoption or governance risk + mitigation)
- (Data or capability risk + mitigation)
- (Dependency that could reorder the phases)

## Decisions needed
- (Owner) -- (decision) -- (target date)

Step 6: Review

Ask the user:

• Which Phase 1 item only reaches a demo, not production? (If any, it does not belong in Now.)
• Where are we deferring platform or governance work that the very first use cases actually need?
• Which phase assumes adoption will just happen? (Name the change-management work or move the item.)
• What is the one dependency that, if it slips, reorders everything?
• Do you want an executive version (outcomes and phases only) or a delivery version (use-case-level detail)?

Revise based on feedback. The roadmap text is finalized before any deck or distribution.

Anti-patterns

Anti-pattern	Why it fails	Do instead
Roadmap of pilots	Many POCs, nothing in production, value never lands	Every phase names a path to production with an owner and a "done" condition
Platform deferred to the end	Each use case re-solves data, eval, access, monitoring	Sequence reusable capability alongside the use cases, starting in Phase 1
Governance as a Phase 3 cleanup	Leaders will not trust or scale AI they cannot audit	Build the approval, monitoring, and ownership model from the first shipped use case
Use cases sequenced, operating model ignored	The work ships, nobody changes how they work, adoption stalls	Carry change management and adoption in every phase, not as an afterthought
Value scored, feasibility ignored	High-value items land in Now before the org can build them safely	Score value x feasibility; low-feasibility high-value items become Later bets the platform unlocks
Date-certain commitments across all phases	Treats directional bets as promises, erodes trust when they slip	Now is committed; Next and Later are intent with triggers to revisit

Output location

Present the roadmap as formatted text in the conversation for the user to copy into their planning doc or feed into /now-next-later-roadmap for a branded deck.