Certance Research · July 2026
Your engineers use AI every day. Adoption keeps climbing, acceptance rates look healthy, and the tooling budget renews itself. Then the board asks what the licences return, and there is no number. That silence has the same cause in nearly every engineering organisation we assess.
Four findings recur in almost every stream we assess.
Most teams recognise at least two of these. The useful question is which one is costing you most right now.
Measurement is nobody's job. Usage dashboards come free with the licences; outcome numbers require deciding what to track, and adoption always outruns that decision. Governance trails for the same reason: agents arrive team by team, review and audit controls arrive later, if at all. No one's engineers failed; adoption outran management. And it carries one real risk: an agent-produced test that incorrectly validates a business scenario, and an auditor finding it first.
Seven dimensions, scored 0 to 5 against evidence, never intentions. A team that merely plans to review AI-generated code scores 1 on governance. You receive per-team scores with the evidence behind them, a leadership decision table with the cost of inaction named, and a 30-day action per team. The engagement takes days and runs on artefacts; it never needs access to your systems.
| Dimension | What it asks |
|---|---|
| Adoption & fluency | How widely and how well AI tools are used: review habits, prompt craft, critical judgement. |
| AI-driven refactoring | Whether AI improved the code that already existed, or only generated new code. |
| Agent creation | What agents are built and deployed, from single-tool helpers to orchestrated pipelines in CI. |
| Quality impact | Measured outcomes: tests created, time to test, defect detection. |
| Process acceleration | Time recovered, tracked rather than remembered. |
| Governance & oversight | Review of AI output before merge, audit trail for agent runs, access boundaries enforced in systems. |
| Direction | How concrete the roadmap is: scoped and resourced, or discussed in a meeting. |
Working out what your organisation actually gets from its AI spend? Certance runs this assessment for engineering teams. If evidence for your regulator is the nearer fire, start with the Coverage Audit.
A structured evaluation of how far AI tooling has been embedded into engineering practice, scored across seven dimensions against evidence rather than intentions. It answers what an organisation actually gets for its AI spend.
Adoption and fluency, AI-driven refactoring, agent creation, quality impact, process acceleration, governance and oversight, and strategic direction. Each dimension is scored 0 to 5, and every score must survive the evidence behind it.
Days rather than months: a structured interview of 30 to 45 minutes per team, verification against artefacts, and a results document with per-team scores, findings, and a 30-day action each. Scores are validated with each team before the report is written.
No. It runs on interviews and artefacts: tool metrics dashboards, repository inventories, CI history. No production access and no customer data.
