The one number I can give you on the spot
There's one stage that prices itself before you've settled anything else: diagnosis. For me that's the AI Readiness Audit — starting at 3,200 PLN net, one to three days of work on-site depending on the company's scale. The price is transparent because the scope is fixed up front: a set format of work and a set output — a report with an implementation map and a return estimate.
The audit is also the cheapest way to avoid overpaying for everything that comes after. Without it, the first implementation gets picked by gut feel. With it, you know which process to support first and roughly what scale of effect it will bring — and that translates directly into a sensible budget.
Why an implementation can't be priced from a rate card
The first project itself is a different matter. Here, "it depends" sounds like a dodge, but it's an honest description. The price of a first implementation depends on several things you can't see before you look closer.
Process complexity is one part of it — supporting a single repeatable task takes different effort than a workflow full of exceptions. Then there's the state of the data: organized and accessible data lets you start right away, while data scattered across spreadsheets, emails and people's heads has to be gathered first. And finally, integrations — a solution wired into an existing ERP system costs more than a standalone tool. A company that quotes a price for a first implementation without looking at any of this is either guessing or has built a buffer into the price for its own uncertainty. That's why, in my case, the first implementation gets priced after a consultation, not off a price list.
One thing about the mechanics is worth knowing. The first pilot implementation is billed in stages, with a decision point after the first week. You don't put the whole budget on the table blind — after the first stage you can see whether the project is heading the right way, and only then does the decision to continue get made.
What actually costs the most in this equation
The most expensive line item in an AI implementation is rarely on the invoice. It's the cost of a bad project — the money and months sunk into a tool nobody uses. I've seen companies spend more on a license for a system used at one-tenth capacity than a dedicated project solving their actual problem would have cost.
On top of that come costs that are easy to miss when you're just looking at the offer: the internal team's time, which has to go in regardless, maintenance and further development after go-live, and finally, dependency on the vendor if the know-how never stays in the company. A cheaper offer that leaves the company with a tool it doesn't know how to run can end up more expensive over a year than a pricier one that leaves it self-sufficient. I've laid out a fuller comparison of these paths — build it yourself, hire a software house, buy an off-the-shelf product — separately.
How to cut the bill in half
For manufacturing SMEs there's a real lever here. The Dig.IT grant covers up to 50% of eligible implementation costs, and building a dedicated solution falls within its mandatory component. On a well-structured project, half the bill goes to the grant. Exactly what the grant funds, what it doesn't, and how to structure a project around it — I break down in the article on Dig.IT's eligible costs.
What's next
If you want to estimate the cost of a first implementation in your own company, the order is simple: start with the diagnosis, which has a transparent price and gives you a map, then price the implementation itself based on that. I describe the full path — from diagnosis, through the first project, to what comes after — on the offer page.
→ See the full path and packages
→ Book a 30-minute consultation
