People are faster with AI. Is the business better off?
A faster draft is a useful start. The harder question is whether the whole job gets done sooner, with less review and a result someone can use.
What the survey actually measures
- 80% of respondents said AI improved their personal productivity.
- 37% attributed some enterprise-level EBIT impact to AI. EBIT means earnings before interest and taxes.
- About 20% said operating costs, including token costs, constrained AI use.
McKinsey surveyed 1,719 people in 97 countries in May–June 2026. These cross-industry findings are self-reported. Read the original survey.
These percentages answer different questions. Their difference is not a failure rate or a measure of lost profit.
Why it matters for commerce
Consider a supplier comparison. A researcher produces it quickly, but the sourcing lead still needs matching quantities, current quotes, freight assumptions and a clear recommendation. Finance reviews the landed cost. The project owner decides what to do next.
If the comparison reaches finance with incompatible assumptions, faster research has simply delivered the problem earlier. If it arrives with the evidence attached and the unresolved questions clearly named, it may help the whole team move. The difference is visible only when we follow the work past the first output.
WIZIUM perspective
Our reading: follow the saved hour
We suggest treating saved time as capacity that still needs a destination. A team might use it to review more products, improve a launch plan, respond to clients sooner or take on another project. Those are different benefits, with different evidence.
An hour saved on drafting is not automatically an hour removed from payroll. A team can become more capable while its salary bill stays the same. That may be a good investment, but the business case should name the additional work the team can now complete.
The same distinction applies to quality. More detailed output may create more review work. A shorter comparison with traceable assumptions may be more useful than a long answer the reviewer has to reconstruct. These are operating hypotheses to test in your workflow, not findings measured in the McKinsey survey.
Track the work from input to decision
Choose one recurring job with a recognizable finish line: a reviewed product shortlist, an approved listing or a supplier comparison ready for a decision. Record the same measures before and during the AI-assisted trial:
- Hands-on time: preparation, prompting, checking and corrections, across everyone involved.
- Elapsed time: how long the job waits between the initial request and an accepted result.
- Review quality: whether the reviewer can accept the first version and which gaps cause another round.
- Full cost: the services used, usage charges, setup and continuing maintenance.
- Useful completion: whether the responsible teammate can take the agreed next action.
Compare jobs of similar difficulty. Keep the acceptance criteria and reviewer consistent. Record changes in scope, missing inputs and unusual cases rather than quietly dropping them. A small trial is a way to learn where work improves; it is not a universal productivity benchmark.
Make one handoff explicit
A useful supplier-comparison handoff might contain the dated source quotes, comparable order quantities, the assumptions behind landed cost, unresolved conditions and the person responsible for the decision. The exact checklist belongs to the team doing the work.
Then ask the reviewer a concrete question: “What did you have to find or redo before you could use this?” That answer identifies a change worth testing. Repeating this check is more informative than collecting screenshots of impressive AI responses.
Try this with your team
Define the finish line before choosing the tool
- Pick one recurring job and write down what an acceptable result contains.
- Name the person who accepts it and the teammate who acts next.
- Track preparation, review, waiting and corrections across comparable examples.
- Decide whether the gain is lower cost, faster completion, better quality or greater capacity.
Scale the workflow when its benefits remain visible after the checking is counted.