As of 27 October 2026 · checked by Oskar Korjus
We do not publish prices for client work. Every build is shaped around your workflows, systems and data. We price the work after we understand it. The first step is a free 30-minute audit call with Karl. Before any work starts, scope, integrations, ownership and cost are in writing.
You may have come here looking for a number. Our site does not show one for client work, and I understand if that is annoying. The reason is below. So is what the missing number costs you, and where to find figures you can plan with. By client work I mean the audit and the builds that follow it.
Why we price the work after we understand it
Our home page says it in one line: "We price the work after we understand it." The text under that line reads: "The audit gives us the workflow, the systems involved, the constraints, and the outcome worth measuring. From there, we define the first useful build and explain the implementation, integrations, ongoing ownership, and cost in writing before anything starts."
Every build we do is shaped around one company's workflows, systems and data. Until we have seen yours, a number on a web page would describe someone else's project. The longer case for working in that order is in what an audit-first AI consultancy is.
Our site used to show a monthly starting figure for client work. We took that figure off the site in September 2026.
What a missing price costs you
Nielsen Norman Group (NN/g) studies how people use websites. In State the Price to Give B2B Sites a Competitive Advantage (1 December 2013), Hoa Loranger of NN/g writes: "In numerous usability studies with business customers, we witness people getting frustrated and leaving sites that don't show prices." The same article says people see companies that hide costs as "evasive and untrustworthy".
That is a real cost for you. You have a budget to plan. We are asking you to talk to us before you see a figure.
The article also says what a company should do about it. When exact costs are unrealistic, it recommends showing prices for typical scenarios or a price range. We have chosen not to do that.
Instead, we do four things:
- If you ask us for a price, you get it in writing once we know enough about the work to give one.
- The audit call with Karl, my co-founder, is free. You leave it with a clear next step. A short summary follows by email.
- Before any work starts, the scope, integrations, ownership and cost are written down.
- If a build is not worth doing, we say so.
What decides the cost of a build
When we price a build, four things move the number:
- how many workflows it covers, and how complex each one is;
- how many of your systems it has to connect;
- the state of your data;
- where the system is hosted.
The audit call is where we first learn about your workflows and the systems involved. Our post on what an AI audit for a business covers explains the two kinds of audit people mean by that phrase. Our page on AI consultancy in Estonia describes the same order: understand the work first, then write down the build, integrations, ongoing ownership and cost before anything starts.
How to plan without our number
Loranger's article makes one more point we agree with: "having even ballpark figures helps in the planning phase." We still do not publish our own. Three of our posts collect dated public figures and a method you can use with any vendor:
- How much does an AI consultant cost? gives dated public rate data for AI consultants from ContractRates and IT Jobs Watch, and compares hourly, project, retainer and outcome pricing.
- AI running costs lists the six lines on an AI bill after launch and the four numbers worth tracking, with a sourced Tallinn wage figure to compare against.
- How to compare AI automation quotes shows how to put quotes side by side: write your own one-page scope, have every vendor price the same unit of work, then check each quote against eight lines.
The last tip comes from NN/g as well. In Show Prices for Common Scenarios (9 April 2006), Jakob Nielsen advised sites that cannot show exact prices to "offer users some representative cases and their prices." You can ask for that yourself. Ask any vendor what a typical case close to yours has cost.
What a number does for a buyer, and what to check in the quotes you get
NN/g names three reasons business buyers need a price. The price tells them which product category they are looking at. It is a key part of comparing candidates. It is also needed for planning, because large purchases take a long time to plan and budget decisions need at least a rough figure.
We do not give you that number in advance, so the quotes you collect from other vendors carry more weight. Our post on comparing quotes points out that three vendors can describe one job in three different shapes: one prices a pilot, one prices a platform and one prices a team. Its fix is to write your own one-page scope and send that same page to every vendor. It then asks each vendor to price one unit of work at today's volume and at three times today's volume.
After that, it reads every proposal for eight things and marks each one present, vague or absent:
- scope, in your words;
- integrations, and who provides access;
- what it costs to run, separately from what it costs to build;
- whose name the vendor accounts sit in;
- a written acceptance test;
- a timeline with milestones, not a duration;
- what happens when scope changes;
- handover and exit.
The post says the pattern of blanks tells you more than the prices do. It also says the cheapest build price and the cheapest total almost never belong to the same vendor.
What this post does not claim
- It does not claim the site has no prices at all. Two fixed-price offers show their prices on their own pages.
- It does not claim we never give a price. If you ask, you get one in writing once we know enough about the work.
- It does not claim that the audit call sets a price.
- It does not claim we know how many buyers leave because there is no number.
kratt is an audit-led AI systems partner in Tallinn: we work out where a company loses time, money or information in its workflows, then build the practical AI, automation or systems integration that fixes the most valuable part.