Contents7
AI saves time: almost every survey says so, and the employees who use it confirm it. Far fewer studies ask the next question: what happens to that time? An hour saved on writing up a meeting appears nowhere in the accounts. It can go into doing the work better, doing more of it, catching one’s breath, or it can vanish into reviewing what the AI produced.
This is the blind spot of many AI projects in a team of 20 to 500 people: the tool gets measured, rarely what the organisation does with what it frees up.
In short
In a survey by BCG, which sells AI consulting, of employees at companies with more than 100 million dollars in revenue, 42% of frontline employees who use AI regularly say they save at least one day a week.
In the same survey, 66% of those employees have little or no guidance on how to use that time.
BCG writes it itself in a note for finance and IT leaders: productivity gains do not reduce cost on their own.
A gain that is not reallocated does not disappear, it scatters. Management’s job is to decide where it goes, without turning it into a threat to anyone.
What the surveys say about time saved
The most quoted figure comes from BCG’s “AI at Work 2026” study; BCG sells AI consulting. It surveyed 11,749 people across 14 markets, all employed by companies with revenue above 100 million dollars: a snapshot of large organisations, not of 50-person teams. It measures a reported gain, not a timed one.
of frontline employees who use AI regularly say they save at least one day a week
of them have little or no guidance on how to use the time saved
of respondents spend more time reviewing and correcting AI
Another survey gives a far more modest order of magnitude. Deloitte, which sells consulting, audit and AI services, had Ipsos survey 25,000 British workers, with no breakdown by employer size: those who use generative AI at work report saving 70 minutes a week, mostly on searching for information, writing emails and summarising. A day on one side, just over an hour on the other: the two figures measure neither the same population nor the same use, and both are self-reported.
The gap between perception and measurement is not a detail. In a study by METR, a research organisation, 16 experienced developers took 19% longer with AI on projects they had known for years, while believing afterwards that they had saved 20%. It is a small sample with early-2025 tools, but the lesson holds: a reported gain is a starting hypothesis, not a result.
Why saved time cuts no cost on its own
Imagine an account manager who saves three hours a week on her meeting notes. Her salary does not change, nor does her role, and her department’s workload stays the same. The company’s costs have not moved by a single euro. What has changed is capacity: three hours that can go somewhere.
Productivity gains do not reduce cost on their own.
The same BCG note adds two instructions worth keeping: reduce and consolidate before you AI-enable, and redesign rather than retrofit. In other words, if a process was badly divided before AI, it stays that way afterwards, only faster.
Without a decision, saved time follows its natural slope.
What saved time becomes without a decision
- The gain
An employee finishes a repetitive task faster.
- The scatter
The time goes into reviewing AI output, messages, finishing touches.
- The blind spot
Nobody knows where it went, or whether it was useful.
- The doubt
Leadership sees no effect and starts questioning the tool.
This is nobody’s fault. Reviewing is necessary, and BCG counts 41% of employees who say they feel a heavier mental load. But time with no destination cannot be managed.
What organisations that decided are doing
Some organisations have made the choice explicitly. EY, which sells consulting and AI services, surveyed 500 US executives of large organisations that invest in AI. Among those seeing productivity gains, 17% have reduced headcount; the others mainly reinvest the gains elsewhere.
Where organisations reinvest their productivity gains
These figures describe large US organisations, not a European SME, and they are self-reported. They still show something useful: headcount reduction is the minority destination for saved time, even where it would be easiest to justify. In Denmark, a study published by the NBER, which links surveys to administrative registers, finds no measurable effect of generative AI on workers’ earnings or hours: employers mostly reorganise tasks.
Jobs do not vanish: they shift. BCG counts 47% of employees who say their role has moved towards steering AI. That is new work, and it deserves to be recognised as such.
Four possible destinations for freed-up time
For a team of 20 to 500 people, we see four legitimate uses of saved time. None is better than the others in the abstract; the mistake is not choosing.
- Absorb growth without hiring in a rush.
If your activity grows, the freed time lets the same team handle more files, and lets you hire at the right moment rather than under pressure.
- Tackle what kept being postponed.
Every team has its list of jobs never done: contacting former clients, documenting a process, cleaning data. That is often where time pays back most.
- Train.
Saved time is the only moment when people can learn without eating into production. Beginners in particular lose the simple tasks that used to be their school.
- Raise quality.
Review better, answer faster, polish a sales proposal. It is less visible than a saving, but measurable if you pick an indicator before starting.
The right choice depends on your bottleneck. If work piles up in front of a single person or a single step, saving time elsewhere achieves nothing: our guide to finding your business bottleneck helps you spot it before deciding.
What we think
The public debate about AI at work swings between two stories: spectacular gains and threatened jobs. The BCG figure suggests a more ordinary and more useful reality: AI frees time, and most organisations have not decided what to do with it. That is neither good nor bad news; it is a management task.
From the field
Measure on a cost line, not on a feeling
The clearest gain we have measured ourselves is not about time but about a bill: on Palmora Property’s translation pipeline, switching provider divided the cost by 60, and 95% of segments come out of the cache after the first pass (project CNT.7). That figure can be checked because it sits on a cost line. Saved time can only be checked with an indicator chosen before the project.
We add one precaution: telling a team that AI will “free up time” without saying what it will be used for leaves everyone imagining the worst. BCG notes that 50% of respondents have no clear governance of work shared between humans and AI. Saying early where the time will go reassures more than any speech about transformation.
Where to start
Before equipping a task, write one line on what the saved time will become and which indicator will prove it. If the task can be solved by a simple rule, without AI, even better: see Automation without AI: when a rule is enough. For an overview of the method, our guide to integrating AI in your business sets out the steps, and our analysis of the real return on AI explains why gains are often slow to show up in the accounts.
If you want an outside person to observe the tasks with your team and estimate what you can realistically expect, that is what our AI Audit Sprint is for. And to remove a re-keying task for good, our n8n automation offer starts with the same question: what will the returned time be used for.
Frequently asked questions
How much time does AI really save?
Surveys give very different orders of magnitude: at least one day a week for 42% of frontline employees who use it regularly according to BCG, in large companies, and 70 minutes a week according to Deloitte in the United Kingdom. These are self-reported gains. For your team, only a before-and-after measurement on a specific task gives a reliable figure.
Should you cut headcount when AI saves time?
That is not what most surveyed organisations do: according to EY, 17% of those seeing gains have reduced headcount, the others reinvest. Work shifts towards reviewing and steering AI, quality or growth. The decision is yours, but it is easier to make once the gain is measured.
How do you measure time saved with AI?
Pick a task, measure its duration and quality for two or three weeks before the tool, then after, with the same people. Also record review time: BCG counts 52% of employees who spend more time correcting AI. Without that baseline, the gain remains an impression.
What should you do with the time AI frees up?
Four uses hold up: absorbing growth, tackling postponed jobs, training, raising quality. What matters is choosing one per team and saying so early, so the time does not scatter and nobody reads it as a threat.
Sources
- Saving 70 minutes a week (new tab)deloitte.com
- 16 experienced developers (new tab)arxiv.org
- NBER (new tab)nber.org
Jim · founder of Brain Plus AI
Also in business ai
Stepping back11 min
AI return on investment: why it is taking so long
The figure of 95% of AI projects with no return went round the world. What it really measures, why the other studies say roughly the same thing, and the practical lessons to draw from them for an SME.
Understand8 min
Business bottlenecks: how to find one and free it
A quote waiting for a signature, a single person who knows how: the bottleneck sets the pace for the whole company. How to spot it, why speeding up elsewhere achieves nothing, and how automation clears it without taking the decision away.
Stepping back8 min
Will AI take young people's first jobs?
No disruption is visible across the labour market as a whole, but a signal appears among young people, in the occupations where AI replaces the task. What the studies say, our view, and the real question for an SME: how to train its juniors tomorrow.