Service 02
AI development lab
A standing place to try ideas without each one needing to become a project with a business case. You bring the question, we build something within weeks that works on your own data, and only then do you decide whether it should go further.
The trouble with one-off projects
Once every experiment needs a quote, an approval and a project plan, only the safe ideas get tried. Precisely the ideas whose outcome you already know. The interesting questions, the ones where nobody knows whether they work, never get a turn because you cannot write a business case for them.
At the same time there is a real chance an idea does not work. That is fine as long as it cost you two weeks. It becomes a problem when you spent three months and a project organisation before finding out.
What the service covers
Ongoing capacity instead of separate assignments. You buy a number of days per month and decide yourself what they go into.
-
An idea list that belongs to you
We keep a shared list of questions worth trying. It grows on its own once people notice something actually happens with what they suggest.
-
Prototypes on your own data
Every experiment runs on your records, not on a demo set. That is the difference between knowing something works and thinking it does.
-
A bar agreed in advance
Before building we write down what counts as success. Without that, every prototype gets talked into a success afterwards.
-
An explicit stop decision
Each experiment ends in one of three ways: build it out to production, shelve it for later, or throw it away. The third is the most common and that is how it should be.
How it runs
- 01
One-day kickoff
We fill the first idea list and pick where to start. There is usually more sitting around than people expect.
- 02
Cycles of two to four weeks
One or two experiments per cycle, each with its own success criterion. Each cycle ends with a short demo to the people it matters to.
- 03
Quarterly review
What was tried, what survived, what it produced. That is also the moment to stop if the lab is not returning enough.
What you end up with
More ideas tested per year at a lower cost per idea, and an organisation that gets used to a failed experiment being information rather than waste.
- Six to twelve tested ideas a year instead of one or two
- Working prototypes, not presentations about what could be built
- Early knowledge of which ideas are hopeless, before they get expensive
- An internal team that learns to judge what is feasible
What it costs
What an engagement costs depends on its scope and on how many systems are involved. I work with a fixed price per phase rather than open-ended billing: you know what a phase costs before it starts, and you decide after each one whether to continue.
I only quote a figure once I know what we are talking about. That takes a half-hour conversation, at no charge.
Whether this fits
This works well when
- there are more ideas than capacity to test them
- you want to experiment more often than per-project budgeting allows
- you are willing to throw away prototypes that do not clear the bar
- someone can free up half a day per cycle
This fits less well when
- you have one specific problem and nothing else: pick AI strategy instead
- every experiment needs formal approval inside the organisation
- a discarded prototype would be seen as wasted money
Questions about the lab
What if nothing comes out of it?
In individual cycles that happens regularly. Across a full quarter it is rare. If it does stay empty, that is a reason to stop and I will say so myself. That is what the quarterly cancellation is for.
Who owns what gets built?
You do. Code, prompts and documentation live in your environment and you can continue without me. I never reuse your data or results for other clients.
Is this not just hiring by the day?
The difference is in the shape. With ordinary contracting you decide up front what gets built and pay for execution. Here the question is open and you pay for finding out, including the answer that something cannot be done. That asks a different expectation from both sides.
Let's discuss how I can help your business grow
A short call is enough to tell whether there is a good fit. No sales pitch, just an honest read on what is possible.