The catch

How much time do you really save with an AI system?

Q: How much time do you really save with an AI system?

As much as is currently leaking away in repetitive work that follows a fixed pattern and is tied to one person. Calculate your biggest bottleneck: frequency per week, time per instance, and who does it. That amount is your ceiling. Then calculate with half to three-quarters of that, because a review step is still needed.

Q: Do I save hours or do I save money?

First hours, and whether those hours become money depends on what you do with the freed-up time. Fill them with work that generates revenue, and it becomes money. Don't fill them, and it's rest. Both are legitimate, but they're different outcomes, and it's better to know upfront which one you want.

Q: How do I know if my process is suitable for automation?

By two things: it recurs often, and it goes roughly the same way every time. If the pattern is missing, there's nothing to codify. If it occurs rarely, the build time will never be recouped. If you're unsure, first count for a week how often it actually happens.

We can't know that, and nobody who promises it upfront knows it either. What you do get is the calculation you can work out yourself, with the places where it usually goes wrong.

DD Saga • 14 July 2026 •6 min

← back to home

Pillar 02 // The catch

time-savings.md

$ calc time-savings --honest

How much time do you really save with an AI system?

We can't know that, and nobody who promises it upfront knows it either. What you save depends on one thing: how much repeat work in your business is tied to one person. That can be calculated, in ten minutes, with your own figures. Below is the calculation, and the reason we don't put a number on it.

DD DataDrift Digital • 14 July 2026 • 6 min

This is, after the pricing question, the most frequently asked question in this market. And it's the question that gets the most nonsense in response, because a number sells better than a calculation.

We won't give you that number. Not out of caution, but because the number we'd name would be about us, not about you. What you do get below is the way to work it out yourself, with the places where the calculation usually goes wrong.

01 / the answerHow much time do you save with an AI system?

As much as is currently leaking away in work that recurs often, follows a fixed way of working, and is tied to one person. If that's six hours a week, that's your ceiling. If it's half an hour, there's nothing to gain. The question isn't how much a system delivers but how much your biggest leak currently costs.

That's a boring answer to an exciting question. It has one advantage: you can check it before you buy anything.

02 / the figuresWhy doesn't anyone name an honest number?

Because the published figures are so far apart that they no longer mean anything. Anyone looking around this market finds payback periods ranging from two months to fourteen months, and investments from a few tens of euros a month to tens of thousands of euros one-off. Both extremes are true, for different businesses, with different bottlenecks.

An average across that spread isn't a prediction. It's a number that happens to lie in the middle of things that have nothing to do with each other. Anyone who presents it to you as your payback period is borrowing credibility from a sample you weren't part of.

A number that fits everyone fits no one.

There's another reason, and it's less pleasant. A supplier who promises time savings is promising something they can't measure and therefore can't deliver on. Ask for the measurement after six months and there's no baseline measurement. That's why we don't put a number in the proposal, and you can hold us to that.

03 / the calculationHow do you calculate your own leak?

Don't take your whole business. Take one process, the process you complain about most. Then four questions:

1How often does it recur? Per week, counted, not estimated. Seven times a week is different from seven times a month.

2How long does it take each time? Including looking things up, switching tasks and picking it back up. That part is consistently forgotten and is often half of it.

3Who does it? You, or someone cheaper? An hour of the owner's time and an hour of an assistant's time aren't the same hour.

4Does it go almost the same way every time? If not, there's little to capture. This is the question that can render the other three worthless.

Multiply the first three and you have a monthly figure. An example with explicit assumptions: six times a week, forty minutes each time, done by someone whose loaded cost is thirty euros an hour. That's four hours a week and roughly five hundred euros a month. That five hundred euros is your upper limit, not your return.

An upper limit, because a system never removes a hundred per cent of it. Work with half to three-quarters, and with a review moment that stays: someone still needs to check that it's correct before it goes out. Anyone counting on a hundred per cent is counting on a system without a human in it, and that's not what we sell.

04 / the dipWhy does it cost more time at the start?

Because the system doesn't know you yet. In the first weeks you correct a lot, and correcting sometimes takes just as long as doing it yourself. On top of that there's a fixed input moment: once a week telling it what's changed about your offer, your target audience or your way of working. Skip that, and the knowledge layer ages, and quality quietly declines.

So don't count on gains from day one, but on a dip first. In our own practice that lasts a few weeks, and it's not a malfunction: it's the period in which your judgement goes into the system. Anyone who doesn't plan for that dip experiences it as failure, when it's actually the investment.

05 / the shapeSo when do the gains actually arrive?

At the point where correcting becomes rarer than producing. That's not a date but a tipping point, and you notice it in something small: you start approving drafts without rewriting them. From there the gains build up without anything new being built.

And pay attention to what you actually get back, because it's often not what you expected. The hours are the visible part. The part business owners mention to us more often is that the work carries on when they're not there. That's not time saved but detachment from your presence, and that's a different kind of value than a tally of hours.

06 / the limitAnd what if you don't save anything?

That happens, and it's predictable. Three situations where the calculation comes out negative:

The process occurs too rarely. Something that takes two hours and recurs once a month is cheaper done by hand than with a system.
Every time is different. If there's no pattern, there's nothing to capture, and your head remains the one doing the work.
Your bottleneck is at the front end. Too few customers isn't a repeat-work problem. A system then speeds up the part that isn't the bottleneck.

If you land on one of those three, the answer isn't a smaller system but no system. We say that in the conversation too, and it saves us both six months.

Frequently asked questions

How much time do you really save with an AI system?+

As much as is currently leaking away in repeat work that follows a fixed way of working and is tied to one person. Work through your biggest bottleneck: frequency per week, time per occurrence, and who does it. That figure is your upper limit. Then count on half to three-quarters of that, because a review moment is still needed.

How long does it take for an AI system to pay for itself?+

That depends entirely on the ratio between your leak and the investment, and the published figures in this market range from two to fourteen months. That spread is too wide to say anything about your situation. The only honest answer comes from your own calculation, not from a market average.

Does an AI system cost more time than it saves at the start?+

Yes, in the first few weeks. The system doesn't know you yet, so you correct a lot, and there's a fixed weekly input moment on top of that. That dip is the period in which your judgement lands in the system. Plan for it as an investment, not as a start-up problem.

Do I save hours or do I save money?+

Hours first, and whether those hours become money depends on what you do with the freed-up time. Fill them with work that does generate revenue, and it becomes money. Don't fill them, and it's rest. Both are legitimate, but they're different outcomes, and it's better to know in advance which one you want.

How do I know if my process is suitable for automation?+

By two things: it recurs often, and it goes almost the same way every time. If the pattern is missing, there's nothing to capture. If it occurs rarely, the build time will never be paid back. If you're unsure, count for a week first how often it really happens.

30 minutes, free, no sales pitch

Want to run that calculation together on your own figures sometime?

Book a conversation

We'll look at where your time is leaking away and tell you honestly whether we can do anything about it. Often the answer is: you can do this yourself, or this leak is too small to build for. Then we'll say so, and you'll have spent thirty minutes getting a clear answer.

→ Book a conversation (30 min)

This text was produced with AI assistance and checked and approved by a human before publication.