Service 01
AI strategy and implementation
Most mid-sized companies do not have an AI problem, they have a choice problem. A hundred things are possible and it is unclear which of them make money. I help you make that choice, then take one use case through to a running solution.
Why AI projects stall
The pattern is predictable. Someone sees an impressive demo, budget appears, a pilot starts, and nine months later there is something that technically works but nobody uses. It rarely fails on the technology. It fails because the chosen application did not solve a real bottleneck, or because nobody owned it once the external party left.
The second pattern is paralysis. So much is said about what could be done that nothing happens. Both are expensive: the first costs money, the second costs time.
What the service covers
Three phases with a decision point after each one. You are not committed to anything before you know what it returns.
-
Use-case scan
We walk through your processes and collect candidates. Not everything that is possible, but everything where time or money demonstrably leaks today. That usually surfaces ten to twenty options.
-
Prioritisation on ROI
Each candidate gets an estimate of return, implementation effort and risk. What ends up on top is rarely the most exciting idea, but it is the one that pays for itself soonest.
-
Proof of concept
We build the top use case into something that works on your own data. Not a mockup, not a demo on a clean dataset, but something a colleague can genuinely use for a week.
-
Implementation and handover
If it works, we make it production-grade: monitoring, error handling, documentation. And someone in your organisation learns to maintain it, so it keeps running once I am gone.
How it runs
- 01
Two-week scan
Conversations with the people doing the work, a look inside the systems, and a ranked list of opportunities. You get that list even if my conclusion is that you should do nothing.
- 02
Two to four weeks proof of concept
Built on your own data, against a success criterion agreed in advance. If it does not clear the bar, we stop. That happens, and it is cheaper than continuing.
- 03
Four to eight weeks implementation
From working prototype to something that survives production, including handover to whoever will maintain it.
What you end up with
One application running that demonstrably saves or earns something, and an evidenced list of what should come next.
- One working AI application in production, not a pilot that lingers
- A prioritised list of the next five to ten use cases
- Someone on your own team who can maintain the solution
- A measurable number: hours saved, lead time cut or conversion gained
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
- you can tell time is leaking but not exactly where
- an earlier AI initiative stalled and you want to know why
- you would rather finish one thing than half-finish five
- there is someone internally who can take it over afterwards
This fits less well when
- you want a supplier who takes the whole thing over and keeps it
- the most important data is not digital anywhere yet
- there is no budget holder who can decide within two weeks
Questions about AI strategy
Do we need our data in order first?
Usually not. That advice sounds sensible but leads to multi-year data programmes with nothing to show. For most first use cases the data is good enough. If it is not, that surfaces within two weeks during the scan, and then you know exactly which part of your data needs work and why.
Do you build on ChatGPT or on custom models?
Whatever solves the problem most cheaply. In practice that often means an existing language model through an API, combined with your own data. Training your own models is rarely economic at this size and I will not propose it when it is not needed.
What if the proof of concept fails?
Then we stop and you do not pay for the implementation phase. That is not a disaster, it is the point of that phase: finding out cheaply whether something works before you put real money behind it. Roughly one in four proofs does not clear the bar.
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.