Right use. Not just use.
Independent consulting on AI adoption, compliance, transformation, governance and strategy, for UK businesses. Putting AI in is the easy part. The work is finding where it earns its place, and saying so when it does not. No software resold, no referral fees, nothing that needs a site visit.
It speeds work up
The four-hour task that does not need judgement all the way through. Reading, sorting, drafting, checking, summarising.
And joins things up
Six systems that have never spoken to each other, answerable in one question instead of four phone calls.
What that actually means
The right use, or none
Where AI earns its place in your business and where it does not. Most of the value is in narrowing that down before anybody builds anything.
The data question, answered
Lawful basis, what leaves your estate, who can see it and how long it is kept. Plainly, with the parts that genuinely need a solicitor marked as such.
A named owner, a rule
Transformation fails on ownership far more often than on technology. Who decides, who maintains it, and what people are allowed to put into it.
One agreed number
What this is meant to change, written down before it starts, so it can be checked honestly afterwards instead of quietly dropped.
Two lenses, one person
Most AI decisions get looked at once, from one side. The board sees a business case and cannot test the technical claims. The technical team sees an architecture and cannot argue the commercial one. The gap between those two views is where the money goes missing, and the data protection question usually falls down the back of it.
Business lens
Cost, saving, risk, ownership, and whether the board will still care about this in six months.
IT lens
Data reality, integration cost, security exposure, and who is running it at three in the morning.
The data question gets answered up front
Privacy is the objection people raise last and worry about most. In one study only three in ten named it as a barrier on a survey form, while more than eight in ten raised it substantively once they were actually in conversation. It is the question that quietly stops projects, and it is usually asked of somebody who cannot answer it.
Not deferred
Lawful basis, what leaves your estate, who can see it, how long it is kept and what you would have to disclose. Answered as part of the work, not deferred to a follow-up from a specialist with their own invoice.
Not overstated
I am not a lawyer and I will not pretend to be one. Part of the job is telling you which parts are routine and which parts genuinely need a solicitor, so you are not paying one to learn basic facts about your own systems.
AI theatre, and the alternative
Anyone can put AI into a business. The skill is knowing where it will make a difference to you specifically, and having the standing to say so when the answer is nowhere near where everybody was looking. Three-quarters of executives now admit their AI strategy is more for show than substance. That is an expensive thing to admit and a cheap thing to avoid.
AI theatre
Buy a platform. Announce it. Then look for a use. Report licences bought as progress. Stop mentioning it around month nine.
The right use
Start from something already costing you money or patience. Ask whether AI beats the alternatives. Sometimes the answer is fewer steps and no software at all, and I will say so.
Things I have actually built
Advice is easier to trust from somebody who builds. The Assurance Register checks a UK job advert against fifty named employment law rules before it goes out, and there are free checks every month. The Compliance Register puts every legal duty a 10 to 100 person business carries on one screen, and there is a sample of it you can click through.
What it costs
One rate, £1,000 a day, with no VAT to add. Every price is fixed before work starts, and any fixed price can be paid in three monthly instalments at no extra cost.
The review
A proper report. Every AI tool already in your business including the unapproved ones, where your data actually sits, what is worth doing, what is not, and the data protection position on each.
The work
Starts at two days: one input, one output, nothing touching a live system. Beyond that it is £1,000 a day, quoted as a fixed number of days before it starts. Never open-ended, never revised upward without a new quote.
On call
One hour a month plus email throughout. Board papers read, vendor quotes checked, contracts looked at before they are signed. If something needs more than that, we agree extra days first.
Straight answers
Is AI expensive?
The AI is not. The mistakes are. Licences cost tens of pounds per person per month. What runs up a large bill is buying a platform before you have a use case, paying for seats nobody opens, and discovering nine months in that the data was never fit for it. My own fees are on the services page in full: a first piece of built work is two days at £2,000, and most builds land between £4,000 and £8,000. The point of paying for a review first is that it costs less than the wrong build.
What does your time cost?
A thousand pounds a day, and there is no VAT to add, so the quoted number is the number invoiced. The review is a fixed £1,250 and comes straight off the first invoice for work commissioned within ninety days. Built work starts at two days, quoted as a fixed number of days before it starts, never open-ended and never revised upward without a new quote. On-call support is £500 a month, cancellable on thirty days notice by either of us. Any fixed price can be paid in three monthly instalments at no extra cost, and there is no travel, no expenses and no disbursements, because everything is remote.
We already pay for Copilot. Have we adopted AI?
Probably not in any way that reaches your accounts. Around 16 percent of UK firms have deployed AI deliberately against a defined purpose. Around 70 percent have some in the building because it arrived bundled with software they already had. The gap between those numbers is where the value sits, and licences will not carry you across it.
Our data is a mess. Should we fix that first?
Everyone's is, including the businesses you assume have it sorted. Preparation is commonly 40 to 60 percent of the effort on any AI project, which is exactly why the answer is not a two-year data programme. Pick one narrow area where the data is good enough, prove the value, and let that fund the tidying up.