AI solutions
AI consulting
Working out the task and the solution: where a model is needed, where search is enough, and where plain code will do.
- Reply to your brief
- within 24 hours
- First call
- 30 minutes, no commitment
- First working version
- in 2 weeks: a prototype on your own data
Process
What happens, and when
The work is split into stages, and each one has a named result. You see it yourself rather than reading about it in a report.
Days 1–3
What the process costs
We start with the process, not the model: what it costs today in hours and money, and which number is meant to move. Some briefs change shape at this point — the pain turns out not to be where the AI was going to go.
You get — The process counted in hours and money
Days 4–7
Options side by side
The task is laid out as options compared in the same units: plain rule-based code, search across your documents, an off-the-shelf subscription service, a model behind an API, an open model on your own server. Each has its own cost to start, its own timeline and its own price for being wrong.
You get — The options weighed in the same units
Week 2
Prototype on your data
We build a prototype on your own data — not a demo on convenient examples, but a run across a sample that includes the hard cases. This is where the real answer appears: the accuracy you need is reachable, or the task runs into incomplete data, which no model fixes. The prototype and the example set stay with you.
You get — A prototype and example set you keep
End of week 2
Estimate for production
We price what it takes to carry this into production: where keys and limits live, who watches spend and answer quality, what has to run on your own hardware if the data cannot leave. It belongs in the estimate rather than arriving as a surprise after the start.
You get — An estimate splitting launch from monthly running
From there
Build or not
The answer is not always “build it”. If it is not worth doing, you hear that with numbers behind it; if it is, you hold an estimate, a prototype and an example set you can take to any team, including your own.
You get — A verdict you can take to any team
What the work covers
What the process costs now
the hours and the money it eats, who is tied up in it, and the figure success will be judged on — counted before any talk of models
Options side by side
from rules in ordinary code to an open-weights model on hardware you own — compared on cost to start, on timeline, on the price of being wrong, and on whether a person stays in the loop
A prototype on your data
a working run over your own sample, including the cases that usually break things; the prototype code and the examples are yours to keep, even if the answer turns out to be “no”
Is the accuracy reachable
the sample shows whether the solution clears the threshold you need and what stands in the way — too little data, data that contradicts itself, or data labelled inconsistently
An estimate for production
the cost to launch and the cost of a month apart, model spend worked out from the prototype, and the work on your side — access, storage, someone on duty
The answer, with numbers
build or not, on which option and in which order — a “not worth it” arrives with the same arithmetic behind it as any other conclusion
What we need from you
- The numbers behind the process, or agreement to gather them in the first days — without them there is nothing to compare the options against
- A sample of your real data, hard cases included, for the prototype to run across
- Someone who can say “we are not doing this” — the work ends in a decision, not in a report
What people usually ask
Contact
Send a description of the task
A reply with the scope, the timeline and a budget estimate comes within 24 hours.
The first call is 30 minutes, with no commitment on your side.
01
You describe the task
Five questions in the form, or a plain email — whichever suits you.
02
We answer within a day
With the scope, the timeline and a budget estimate, based on what you told us.
03
We talk for 30 minutes
To clear up whatever is unclear. It commits you to nothing.