Home / Services / AI Consulting and Strategy
Consulting · Architecture · Fractional CTO
AI Consulting and Strategy
Decisions, before the money is spent
Architecture, technical audits and a roadmap for teams adopting AI. I advise on the exact systems I have already built and run, not on a survey of the field.
Vendor-neutral
no commission from anyone I recommend
Operator
I run the systems I advise on
Written
you keep the document, whatever you decide
Days
not a six-week engagement to say what fits
Pricing
What advice costs.
Fixed per engagement. I take no commission from any vendor I recommend, which is the whole basis on which this is worth paying for.
Architecture review
On requestfixed
- Your AI system reviewed: architecture, prompts, retrieval, evaluation
- Where it will break, and where it is quietly wasting money
- Cost per request modelled and compared against alternatives
- Findings ranked by impact, with the fix for each
- A written report you own
Usually pays for itself on the inference bill alone.
Request this ↗AI roadmap
On requestfixed
- Every process that could use AI, assessed honestly
- What to build, what to buy, and what to leave alone
- Sequenced by payback, with costs against each
- Build-versus-buy on the specific tools you are considering
- Presented to your team or your board
Portable by design. If you execute it with someone else, the roadmap still did its job.
Request this ↗Fractional CTO
On requestper month
- A technical decision-maker on call, without the salary
- Vendor proposals and contracts reviewed before you sign
- Architecture decisions and hiring input
- Your team's technical questions answered
- Board and investor sessions when they are needed
For a company with budget and no senior technical person to spend it well.
Request this ↗Add-ons — buy with any package or on their own
Team enablement
On requestfixed
- Workshops for your developers on the tools you have chosen
- Working code against your own systems, not slide examples
- Review of what they build in the weeks after
- A written reference they keep
So the capability stays after I leave, which is the point.
Request this ↗Compare
Side by side.
Whether you are checking something built, deciding what to build, or want the decisions owned continuously.
| What you get | Architecture review | AI roadmap | Fractional CTO | Team enablement |
|---|---|---|---|---|
| Reviews a system you already run | Included | Not included | Ongoing | Not included |
| Decides what to build, buy or drop | Not included | Included | Included | Not included |
| Cost per request modelled | Included | Included | Included | Not included |
| Written report you keep | Included | Included | Not included | Included |
| Presented to your team or board | On request | Included | Included | Included |
| Vendor proposals reviewed before signing | Not included | Included | Included | Not included |
| Hands-on with your developers | Not included | Not included | Sometimes | Included |
| Capability stays after I leave | Not included | Not included | Not included | Included |
| No commission, from anyone | Included | Included | Included | Included |
The review usually pays for itself on the inference bill alone, which makes it the cheapest way to find out whether the rest is worth buying.
Before we start
What your business needs in place.
Advice is only worth its price when these are true.
The decision-maker in the room
Advice relayed second-hand arrives without its reasoning and gets overruled by whoever spoke last. If the person who signs cannot attend, the engagement is worth less than it costs.
Honesty about what has already been decided
If a platform has been chosen and the review is for cover, say so — I will decline rather than write it. Where the decision is genuinely open, that is where this is worth the money.
Access to what is being reviewed
Code, architecture docs, the actual bill, or the proposal on the table. A review conducted on descriptions is a conversation, not an audit, and should not be priced as one.
Willingness to hear 'do not build this'
It is the most common recommendation and the most valuable one. If it is unwelcome, the report will be worth exactly nothing to you.
Terms
How the money and the risk are split.
No retainer to start a conversation, and no invoice for work that never got delivered.
Half to start, half on delivery
Balance falls due when the work is done and handed over. Not delivered, not charged.
Fixed price, not an estimate
The number holds unless you change the scope. If I misjudged it, that is mine.
The clock starts at handover
From the day I have everything on the list, not from the day you pay.
One round of revisions included
Then hourly, quoted before anything starts. No surprise invoices.
You own everything at the end
Code, accounts, assets and credentials. Nothing is held hostage.
Proof
Advice from someone who runs the systems.
The recommendation always comes with an implementation cost I would have to honour myself.
OurGrowth
I built and run a multi-tenant AI platform with 70 skills and an orchestrator. The cost and quality trade-offs I advise on are ones I pay for.
our-growth.com ↗AI Glossary and MCP
A published catalogue of Claude skills as an npm package and an MCP server — the tooling, not just an opinion about it.
Banking, government, telecom
Eight years delivering systems where a wrong architectural call is expensive, which is the experience an AI review actually draws on.
No commission, from anyone
Not a claim about ethics — a structural fact that is the whole reason this advice is worth paying for.
What you get
What you are left holding.
Written to be useful even if you execute it with someone else.
A written findings report
Ranked by impact, each with the fix and what it costs to act on. Yours to keep.
Build, buy or leave alone
Answered per option with numbers, not adjectives — including the options that are not mine.
Your costs, modelled
Per request at your real volume, compared against the alternatives on the table.
Your team, further along
Workshops on your own systems and review of what they build after, so the capability stays.
What you send
The handover, in one list.
Send it in one go and the clock starts.
- What you are trying to change, in business terms
- The current stack and where AI is meant to sit in it
- The proposal, vendor quote or plan under consideration
- Access to the codebase or architecture docs for a review
- Your current inference or licence spend, if any
- Who owns the work after I leave
How it runs
Six steps, and you see each one.
-
01
Brief
An hour on what you are actually trying to change, and what has already been decided.
-
02
Look
Code, architecture, bills and proposals. No opinion before evidence.
-
03
Model
Costs and alternatives compared with numbers rather than adjectives.
-
04
Write
Findings, ranked, each with a recommendation and what it costs to act on.
-
05
Present
To the team, the board, or both. Questions answered in the room.
-
06
Follow up
One session a month later, when the decisions have met reality.
Scope
What this is not, and who it is not for.
These are quoted separately, so the price above stays a price.
- Building anything — that is a separate service, quoted separately
- Vendor licences and subscriptions
- Recruitment and placement fees
- Legal or regulatory sign-off
- Training data or dataset acquisition
- Acting as your named technical officer for regulatory purposes
Right fit, wrong fit
Saves a call that would end in no.
A good fit
- You are about to spend real money and want a second opinion
- You have an AI system and the bill is climbing
- You need build-versus-buy answered with numbers
- You have budget and no senior technical person
- The decision is genuinely still open
Not a fit
- You want a report supporting a decision already made
- Nobody senior can attend
- You want the deck rather than the thinking
- You need someone to build it — ask for the build instead
After launch
After the report.
A roadmap ages. Providers ship, prices fall, and the build-versus-buy answer that was right in March is wrong by September. The fractional arrangement is for companies that would rather have that tracked than rediscovered.
Questions
The ones people actually ask.
Do you take commission from vendors?
No, from anyone, ever. It is the whole basis on which this advice is worth paying for, and the moment it were not true the advice would be worthless.
What if you recommend we do nothing?
Then that is the finding, and it is usually the most valuable one. Most AI spending in this market goes on things that a database query or a better process would have solved for a fraction of the price.
How is this different from a big consultancy?
I run the systems I advise on and I write the code. That means the recommendation comes with an implementation cost I would have to honour myself, which is a strong discipline against recommending things that sound good.
Can you also build what you recommend?
Yes, and it is quoted separately and later, so the advice is never a sales route. If you would rather someone else builds it, the roadmap is written to be handed over.
How quickly can you start?
A review is usually days rather than weeks — the value is in answering before the money is committed, and an answer that arrives after the contract is signed is just an expensive opinion.
Tell me what you have, and I will tell you what it takes.
The app, the market and the deadline is enough to get a straight answer on scope, price and timing — usually the same day.