Irish organisations have invested heavily in Microsoft Copilot licences. Most cannot point to measurable productivity gains. The reason is rarely the tool — it is the absence of a diagnostic before deployment.
Microsoft Copilot adoption across Irish organisations has followed a consistent pattern: licences purchased at scale, a period of initial enthusiasm, then a plateau where measured productivity improvement is difficult to identify. The investment is real. The return is not.
This is not a problem unique to Ireland, and it is not a problem with Copilot specifically. It is the predictable outcome of deploying AI tools without first diagnosing the workflows they are meant to improve.
Why the ROI gap exists
Copilot is a powerful tool for specific tasks: drafting, summarising, searching across documents, generating first-pass content. It is not a general productivity multiplier that works equally well across all working patterns. Its value depends almost entirely on how the organisation using it actually works — which meetings are being held, what documents are being produced, how information flows between people, and where the genuine bottlenecks in productivity sit.
Most Copilot deployments skip the diagnostic. Licences are distributed, training sessions are delivered, and usage metrics are tracked. What is not measured is whether the workflows where Copilot is being used are the ones where it creates most leverage — or whether the AI is being used as a layer on top of working patterns that were already inefficient.
The result is a tool that is used, but whose productivity impact is dispersed across low-value tasks rather than concentrated where it matters.
Check the work around the tool
Look at where Copilot is used, what people still have to check and whether the end-to-end task is any faster. A high volume of generated summaries does not establish that meetings have improved. Calendar metadata alone cannot tell you whether a meeting produced useful decisions. If people say Copilot gives incomplete answers, check what it could reach: why Copilot cannot find company information covers access, search scope and indexing.
An illustrative example
Suppose a team uses Copilot to summarise a recurring meeting. If the meeting has no clear purpose, producing a quicker summary may save some administration while leaving the larger issue untouched. Review whether the meeting is needed, then measure the time required for the remaining work. This is a hypothetical example, not a client result or a benchmark.
What good Copilot governance looks like
Productive Copilot deployment requires three things that most organisations skip:
First, a baseline diagnostic of how work actually flows before AI is introduced. Without a baseline, there is no way to measure improvement.
Second, targeted deployment — identifying the specific use cases and workflows where Copilot creates material leverage, and focusing adoption there before expanding.
Third, ongoing measurement. Copilot produces usage data. That data needs to be interpreted in the context of productivity outcomes, not treated as a proxy for them.
Getting the investment to work
If your organisation has Copilot licences but cannot demonstrate a productivity return, the answer is not a different AI tool. It is a diagnostic of why the current one is not working. The diagnostic is almost always faster and cheaper than the original licence cost — and produces a roadmap that makes subsequent AI investment defensible.
Contact Acuity AI Advisory to discuss a Cognitive Mirror diagnostic for your leadership team.