AI Adoption Programme
The enterprise-wide programme in this guide — streams, gates, training, workshops and pair working — delivered by senior practitioners, not slideware.
ICAN Consultancy · Executive Field Guide
From principles to a 90-day plan
Leading AI Transformation.
A practical playbook for executives who want AI adoption that is real, measured and enterprise-wide — not another round of pilots. The operating model, the maturity ladder, the governance and the first 90 days.
3Adoption streams, run in parallel
5Maturity stages, evidence-gated
90Days to visible wins and real foundations
In this guide
- Four principles that guide every adoption decision
- Three streams — why one programme must not pretend every team is the same
- The maturity ladder — five stages, each one earning the next
- People & capability — training, workshops, pair working, champions
- Governance & measurement — stage gates and metrics that can’t be gamed
- The first 90 days — visible wins and real foundations
The AI Augmentation Maturity Model (AAMM)
Five stages. Each one earns the next.
Maturity is measured objectively and teams advance through evidence-based stage gates. You only progress when the data proves you’re ready — no theatre, no vanity metrics.
| Stage | What it looks like | Move on when… |
|---|---|---|
| 0 · Aware | Ad-hoc, unmanaged AI use. Policy, approved tooling and baselines being established. | Everyone has safe, approved access; a usage policy is live; baseline metrics are captured. |
| 1 · Assisted | AI assists individuals on discrete, transactional tasks — drafting, querying, generating — with guardrails. | Most of the team has working access and uses AI daily, not just the enthusiasts. |
| 2 · Augmented | Best practice is embedded in individual workflows: shared configuration, prompt and context libraries, AI-authored output reviewed and shipped by humans. | Configurations are in place and AI-assisted work lands cleanly on a weekly basis, with measurable productivity gains. |
| 3 · Agentic | Standardised team workflows and playbooks. AI executes isolated, well-bounded flows end-to-end — triggered by humans. | At least one flow runs reliably end-to-end with AI doing the bulk of the work, and the practice is shared by the team — not one champion. |
| 4 · Autonomous | AI executes scoped workflows on automatic triggers — events, schedules, thresholds — within defined guardrails: bounded blast radius, override paths, humans in the loop. People review outcomes, not every step. | Stable when automated workflows deliver direct outcomes — e.g. a triggered audit finds an issue, drafts the fix, passes regression, and a human approves the change into production. |
The ladder is climbed one rung at a time. A team that leaps is a team you will be rolling back.
Failure modes
Five ways this goes wrong — and the early signals.
Every stalled AI programme we’ve been called into failed in one of five ways. Each has a tell you can spot in the first quarter.
| Pitfall | What it looks like | The early signal |
|---|---|---|
| Tool-first thinking | Licences bought before workflows are understood; adoption measured in seats, not changed work. | The programme’s main artefact is a procurement sheet, not a playbook. |
| Pilot purgatory | Impressive demos that never touch production work. Twelve pilots, zero standard practices. | No pilot has a named owner or a date to become the default way of working. |
| Single-champion adoption | One enthusiast per team carries everything; progress evaporates when they’re on holiday. | Usage data shows one heavy user per team and a long tail of near-zero. |
| Skipping stages | Automation deployed on top of habits that don’t exist; brittle wins that collapse under load. | A team “at Stage 3” can’t show a Stage 2 practice that survived a normal week. |
| Vanity metrics | Dashboards full of active users and prompts sent; nothing about cycle time, throughput or quality. | Nobody can answer “what business result improved?” without changing the subject. |
The common thread
All five are governance failures, not technology failures. The model in this guide — streams, gates, evidence, named owners — exists precisely to make each of these visible early, while they’re still cheap to fix.
The honest question
If your current AI initiative stopped tomorrow, would any team notice within a week? If the answer is no, you have a pilot programme wearing a transformation’s clothes — and it’s worth finding out why.
Momentum without results is theatre. Results without momentum is luck. You need to see both.
Proof
This model, in production.
A three-person ICAN team ran this exact programme for a 120-person blockchain organisation — engineering and business functions alike. Strategy, roadmap, tooling, training, pair working: the full journey. Where internal teams had the capability, we set direction and stepped back; where they needed a working example, we shipped it ourselves.
Read the full case study100%AI adoption across the organisation
+30%Productivity increase in six months
70%Of teams at AAMM Stage 3 in six months
Read the whole playbook.
The four principles, the three streams, the full maturity ladder, people and capability, governance, measurement and the first 90 days — the full 12-page guide, free to read and share.
Download Leading AI Transformation — PDF, 12 pages, 83 KB
You are welcome to share this guide in full, with attribution.
Where ICAN comes in
From playbook to programme.
AAMM assessment
An objective read of where every team sits on the maturity ladder today, and an evidence-based plan mapped to the stage gates.
Agentic AI pipelines
We train your teams to build and operate agentic pipelines that automate delivery and operations end-to-end — to your security and governance standards.
No form, no gate
Book an AI readiness conversation.
Tell us where you are today. We’ll bring an objective view of your AI maturity and a pragmatic plan to move forward. No obligation, no jargon.