In this market, prices are rare. Most providers work with a free intake followed by a quote, which means you only find out if you're in the right range after two conversations. We'd rather put the numbers here, even if that means some people drop out before they call.

That's not transparency as a virtue. It's filtering: if this is far outside your range, neither of us gains anything from a first meeting.

01 / the answerWhat does an AI system cost?

Three amounts for the build, depending on the size of the bottleneck: €2,500 for one defined process, €4,500 if multiple systems or an approval step are added, €7,500 if it touches a core system or requires validation against your own history. On top of that, management: €295, €395 or €595 per month.

All amounts exclude VAT. There's no quote process in between: after one conversation in which we map out the process, the class is fixed and so is the price.

02 / the classificationWhy three amounts and not packages?

Because a package assumes we know what you need. With a bottleneck-based approach it's the other way around: you start with the work that costs the most, and its size determines the price. Three questions determine which class you fall into, and they're questions about your situation, not about our services.

You'll hear those three questions again in the first conversation, framed as process questions. They're not there to push you into a class, but because without the answers we can't quote a price we can actually deliver on.

The price hangs on your bottleneck, not on a package you choose.

03 / the monthly costWhat do you pay monthly, and for what?

For management, and that's different from rent. With rent, your access stops the moment you stop paying. Here, the maintenance stops: your accounts, your data and your documented way of working stay yours and keep existing. What falls away is us keeping it running.

That maintenance covers work you don't see as long as things go well. A supplier changes their connection and something needs adjusting. Your offering changes and the instructions behind it need to follow. A better model comes out and the tuning gets recalibrated. And someone checks whether it's still running, which for a chain touching multiple systems isn't an upsell but a condition.

That's why management isn't optional for the two larger classes. For the smallest form you may skip it, and then we also say what that means: if something breaks, that's a new job, not a warranty case.

04 / the catchWhat's not included?

Three things, and they're listed here because otherwise you'd find out afterward.

• Model usage runs on your own account. We set up the connection with the AI provider in your name and keep an eye on usage, but the bill goes to you. That's deliberate: this way you see what it costs and there's no markup from us in between. For most businesses this is tens to a few dozen euros per month, and it scales with usage.
• External data sources are billed at cost. If a job needs data from outside, we pass that through without margin, with a monthly cap you approve in advance. We don't fold that variable cost into a fixed price, because then we'd be carrying a risk we'd have to pass on to you anyway.
• Work outside the agreed scope costs €125 per hour. That's the only place where we work with an hourly rate. Within scope it's a fixed price with a measurable delivery criterion.

05 / the marketHow does this compare to what others charge?

Look around in the Netherlands and you'll find three price tiers. At the bottom, one-off jobs from around eight hundred euros: one workflow, built and delivered, usually without management. At the top, implementation projects from eight thousand to fourteen thousand euros, often with a minimum number of employees before you even qualify. In between, providers publish ranges of five thousand to twenty-five thousand euros, where the top end belongs to bigger companies than the ones we serve.

We sit in the middle, and that's a deliberate choice. Above the bottom, because we don't sell a loose workflow but a working chain with management and a delivery criterion. Below the agency projects, because we solve one bottleneck and don't sell a program. That's not a discount, it's a different unit.

For us, the revenue isn't in the first bottleneck but in the second and third. That's also the most honest reason to start small: if the first build doesn't deliver anything, there's no second one, and then we've both lost very little.

06 / the limitWhen is this too expensive for you?

When your leak is smaller than the bill. Do the math: how often does the work recur, how long does it take each time, and who does it. If you land on two hundred euros a month in wasted time, you won't earn back a build of two thousand five hundred euros plus management, and you shouldn't do this.

There's another limit that's less about money. If your bottleneck sits on the front end — too few customers rather than too much manual work — none of these three amounts will fix that. Then a system is the most expensive way to leave that problem in place.

In both cases we say so in the conversation. That's not noble, it's practical: a client who doesn't earn it back is a dissatisfied client six months from now.

07 / the alternativeWhat does it cost if you don't do it this way?

The question of what an AI system costs only makes sense next to a second question: what does the alternative cost? There are three, and all three have a price.

• Build it yourself. Technically possible. The tools are open source, the documentation is public, and anyone willing to try it needs only a few hundred euros a year in server costs. So the cost isn't really in money but in attention: someone in your business has to build it, understand it, and keep maintaining it when something breaks on a Tuesday morning. Budget several weeks of work for the first chain, and after that a few hours a month nobody planned for. If you can spare that, by all means do it yourself. We mean that.
• Hire an agency for a single job. For a standalone automation or integration, most providers charge between €750 and €5,000, depending on how many systems need to connect. That's a fair price for real work. Watch two things. Loose jobs stack up: three integrations at three different moments means starting up three times, explaining three times and billing three times. And what you're buying is the connection, not the maintenance — that comes back the moment something changes in one of the surrounding systems.
• Do nothing. The cheapest option on the invoice and usually the most expensive one in practice. Take six hours a week of manual work: receipts, preparing files, typing the same answer again. Convert that into what an hour of your time or your employee's time costs and you get a yearly figure. That figure appears on no invoice and in no bookkeeping, and that's exactly why it's so easy to leave it alone.

Note the difference from the comparison above. In 05 / the market we place ourselves next to other AI builders; here we place building next to not-building. What we sell sits between those three: you're not buying someone's time to build it, but the result it produces, with documentation so you can continue on your own later. The amounts are on our pricing page.

08 / the payback periodWhen do you earn it back?

A worked example with the assumptions included, so you can swap in your own numbers.

Say: an S-build of €2,500 one-off, plus €295 per month management. The bottleneck is six hours of manual work per week. At an internal hourly rate of €75, that's €450 per week, roughly €1,950 per month in time spent on work that involves no decision-making by anyone.

Two things usually make a calculation like this too rosy. So we build them in right away.

Which makes the math short: €2,500 divided by €875 is roughly three months. In year one that's €6,040 in costs (build plus twelve months of management) against more than €14,000 in recovered time. From year two onward the build cost drops away and you're left with roughly €10,500 a year against €3,540 in management.

You earn it back with the hours you net, not with the hours you save gross.

Put the honest counterpart next to that immediately. If those six hours a week don't get converted into something that generates money — more customers, faster delivery, fewer errors — you earn nothing back. You've gained time you spend on something else. That can be a perfectly good reason, but then don't call it an investment.

And if you don't know which six hours it is: that's what the free diagnostic conversation is for. Thirty minutes, no obligation, and sometimes the outcome is that you're better off cancelling two subscriptions and changing nothing further.

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