You Saved 40 Hours With AI. Congratulations. Was the Work Any Good?
Imagine three agencies reviewing their AI work. One claims to have saved 40 hours in a month. Another counts a hundred campaign ideas produced in an afternoon. A third reports spending half a million tokens on brand claims. These are hypothetical examples, not measured Mirel customer results.
Fine. Was any of the work good?
If a human team spent a week producing a hundred mediocre claims, nobody would applaud the number of hours involved. If it produced the same list in an hour, the list would still be mediocre. The speed might make the waste cheaper. It wouldn't make the result better.
Yet agencies routinely discuss AI as if the resources it consumes or saves were proof of value.
What the agency expects from AI
The GWA KI-Whitepaper 2026 (page 15, question 6) asked 78 leaders from member agencies where they saw AI's single biggest opportunity. Efficiency led with 57.7%. New services and products followed at 28.2%. Quality was selected by 7.7%.
This was a non-representative self-report survey asking for the single biggest opportunity. Respondents choosing efficiency may also care about quality. Among this group, operational gains were the most common first answer.
For administration, reporting and repetitive internal work, that instinct is sensible. If a task needs to be done accurately and nobody benefits from spending longer on it, make it faster.
Client-facing strategic work is different. Its value comes from better understanding, more useful choices and better creative direction. The best outcome might take less time. It might also require spending the time saved on a question nobody had thought to ask.
An agency that produces a weak strategy in 20 minutes instead of two hours has improved its production speed. It has not necessarily improved its work.
The wrong metric can make the work worse
Agencies have seen this movie before. Introduce a resource-allocation system designed to maximise billability and the spreadsheet may look healthier. Every person has a slot. Every project gets its allotted time. The best strategist for a problem is unavailable, so someone else takes it. A question that needs an uninterrupted afternoon gets 43 minutes between meetings.
The system becomes more efficient according to the metric it was designed to improve. The client may receive worse work.
AI creates a similar risk. Measure the number of briefs produced and people will produce briefs. Measure time saved and they will optimise for speed. Measure token consumption and someone will find a way to spend more tokens.
None of those numbers tells you whether a creative team received a clearer direction.
Start with what people say about the work
One sign of useful AI is surprisingly ordinary: someone says, 'We would never have thought of that.'
The useful contribution is helping a team see a pattern, question an assumption or explore a strategic possibility that would otherwise have been missed.
Those moments are difficult to count. They are still worth paying attention to. The objective is to make the work itself more interesting, rigorous and useful. Impressive demonstrations only matter if that benefit survives into an assignment.
A sensible measurement approach therefore looks at three things.
Is the work getting better?
Ask whether the strategic recommendation is clear, supported by appropriate evidence and capable of producing useful creative work. Compare briefs using the same criteria for good strategy before and after a change in workflow. Ask creatives whether the brief helps them make choices and where they still need clarification.
You can also examine whether a team has considered genuinely different routes, identified the assumptions behind them and made an explicit choice. More options are not automatically better. Better decisions are.
Is the process becoming less wasteful?
Look for less senior firefighting, fewer brief rewrites caused by unresolved strategic questions and fewer decisions reopened after creative work has started.
These indicators need interpretation. A senior strategist challenging a weak direction early may be evidence of a healthy process. The same strategist repeatedly rescuing a project two days before presentation probably isn't.
Measure the difference between necessary challenge and preventable firefighting.
Are clients and creative teams benefiting?
A stronger strategic process should eventually show up in the work and relationships around it. Creative teams may reach a useful idea earlier. Clients may understand the trade-offs behind a recommendation. Smaller accounts may receive proper strategic attention instead of whatever thinking someone can squeeze into Friday afternoon. When clients bring their own AI-assisted analysis, ask whether the agency helped them test its assumptions and make a better choice.
Commercial outcomes matter too, but attribution is difficult. A campaign's performance depends on creative execution, media, market conditions and many other variables. It would be dishonest to credit an AI workflow for a revenue increase without stronger evidence.
A better question for the next AI review
Don't stop tracking time. It is still a cost, and agencies need to understand costs.
Just don't confuse it with the purpose of the work.
At the next meeting about AI adoption, ask what changed in the quality of the decisions people made. Ask which project avoided a predictable mess, and whether successful projects are hiding broken processes. Ask whether a creative team received a better brief. Ask whether a strategist found a new way to frame the problem.
If nobody can answer, the agency may have bought itself a faster way to do the same work. That's useful. It is not much of an ambition.
Apply the same quality criteria to strategic work supported by Mirel before counting any time saving as a benefit.