AI Strategy
AI implementation roadmap: a realistic twelve months
A twelve-month AI implementation roadmap for mid-market companies: what happens in each quarter, what gates each stage, and the signals that you are moving too fast.
Miguel Torres, Founder, Merjora · Published 26 August 2026 · Updated 3 September 2026
In short
A realistic first-year AI roadmap runs discovery and baseline measurement in the first six to eight weeks, a single reversible production workflow by the end of the first quarter, measured value and a second workflow in the second quarter, capability consolidation — evaluation, monitoring, governance — in the third, and portfolio expansion in the fourth. The gate between each stage is measured value against the pre-agreed baseline, not delivery of the previous phase.
Key takeaways
- One workflow in production beats four in pilot.
- The gate is measured value, not project completion.
- Quarter three is for consolidation; skipping it is why programmes stall in year two.
- Governance work belongs in quarter one, not at the compliance review before launch.
Quarter by quarter
| Quarter | Objective | Exit gate |
|---|---|---|
| Q1 | Discovery, baseline, first workflow live with human review | Baseline captured; workflow running with agreed quality threshold |
| Q2 | Measure realized value; deliver second workflow | Realized value reported against the original range |
| Q3 | Consolidate: evaluation, monitoring, governance, review reduction | Review burden reduced; monitoring in place |
| Q4 | Expand portfolio; re-score opportunities and readiness | Refreshed portfolio with updated confidence levels |
Signals you are moving too fast
- Several workflows are in pilot and none has a measured baseline
- Nobody can state the current quality threshold for a live workflow
- Review time per item is unknown
- Governance classification is still 'in progress' after go-live
Signals you are moving too slowly
- Quarter two ends with no production workflow
- Discovery has produced a long list and no ranked shortlist
- Decisions wait on a platform selection that no opportunity actually requires
Frequently asked questions
- How long before AI delivers measurable value?
- For a narrow, well-chosen workflow with digital inputs, measurable value in the first quarter is realistic. Programmes that begin with a platform selection or a data-warehouse project typically see nothing measurable in year one.
- Should we run a pilot first?
- Run a reversible production workflow with human review instead. Pilots without promotion criteria tend to become permanent.
Where would this apply in your business?
Merjora maps your workflows, scores each opportunity and quantifies the likely value range before you spend budget.
Discover your AI opportunitiesRelated reading
Editorial standard. Merjora publishes analysis, frameworks and publicly documented examples. We do not publish invented statistics, unattributed benchmarks or unverified customer stories.