A UK supplier is now advertising a custom GPT for your business from £995. That’s good news, and not for the reason you might think. A visible number turns a fuzzy decision into a clear one. Instead of asking “how much should AI cost me”, you can finally ask the better question: what does £995 actually buy, and where does it stop being enough?
So let’s do that properly, from a buyer’s chair: a straight look at custom GPT pricing, what you get at £995, what you get further up, and how to tell which one your business actually needs. No vendor knocking, just the facts.
What £995 genuinely gets you
At the £995 end of the market, you’re buying a single-purpose custom GPT. You upload a stack of documents, the supplier wires up an assistant that reads them, and you’ve got something trained on your material for one job. For a narrow, repeatable task, like drafting first-pass marketing copy in roughly your style, that can be decent value. If that’s the whole job, a £995 GPT might be all you need, and in all honesty I’d tell you so rather than sell you something bigger.
Where it gets interesting is what isn’t in the box.
Where £995 quietly runs out
The limits of a cheap custom GPT don’t show up on day one. They show up around week eight, and they tend to be the same ones every time:
- No real extraction. You only get out what you thought to put in. A GPT trained on your uploads knows your documents. It doesn’t know the things you’ve never written down: the way you actually talk a nervous client off the ledge, the reason you do step three before step two, the story you tell on every sales call that closes the deal.
- No audience layer, no stories or frameworks. When the model doesn’t have a real example to hand, it does what every model does in a vacuum. It fills the gap with something plausible. How many times has AI invented a case study or a statistic just to make a paragraph sound finished?
- No governance, versioning, or change log. Nothing records what’s true, what changed, or when. There’s no way to correct the system once and trust the correction sticks.
- No maintenance. It’s frozen on the day it’s built. Your pricing moves, you retire a service, you take on two new people, and the GPT carries on quoting last quarter as if nothing happened.
Put those together and, more often than not, you get voice drift inside about 60 days. It starts on-brand and slowly slides back to sounding like everyone else. Without that grounding underneath it, your AI is a temp on its first day. It’ll do something, but probably not quite what you wanted.
This is the part worth slowing down on
A £995 custom GPT and a £4,995 install are not the same product at two different prices. They’re different things. One is a single chatbot pointed at some documents. The other is a system that every AI tool in your business reads from. Confusing the two is where most of the disappointment with business AI actually comes from.
What the next tier up actually is
Further up the market you’re not buying a bigger chatbot. You’re buying an AI Context Layer. We call ours the Firecore. Same thing, different name, except the branded version has a method and an owner attached to it.
It starts with extraction, not upload. The Kindling Method is a structured session, around two and a half hours on a video call, that pulls the voice, stories, frameworks and working knowledge out of the founder’s head and onto the page. The whole point is to capture what you’d never think to type into a chatbot on your own.
What gets built from that is a governed reference layer: four layers and fifteen modules covering your voice, profile, business, audience, stories, frameworks and goals. It’s versioned, it has a change log and a change-request process, and it’s read-only by design, so nothing changes without a record. Every AI tool you use then points at the same single source of truth instead of each one guessing on its own.
For £4,995 that’s the install, including a live install day, a fast-ingest first month where the system loads your existing notes and assets, a self-archiving workspace, training, and a Personal Strategy Manager built in. Keeping it accurate over time is the optional Firecore Stewardship Plan at £1,497 a month, and that’s governance rather than maintenance: weekly curation, a monthly strategy sync, ongoing market intelligence, and a quarterly onsite audit.
It also stays yours. Everything lives in your own accounts and subscriptions, and nothing passes through our servers. If you ever stop the Stewardship Plan, the whole system stays in your workspace, fully working, frozen at its last good state. That’s a partnership, not a hostage situation.
So which one do you need?
The honest test is scope, not budget.
- If you want one narrow task handled, and you’re happy to rebuild it by hand whenever your business changes, a £995 custom GPT may be exactly right. For a sole trader or a single tight use case, a full install is overkill, and you shouldn’t pay for one.
- If your business runs largely on what’s in the founder’s head, and you want every tool, every draft and every assistant to be accurate and on-brand without you checking each one by hand, that’s an install, not a chatbot.
Most UK service businesses with somewhere between 5 and 50 staff sit firmly in the second camp, buy the first option, then wonder why the output still needs babysitting. British Chambers of Commerce research in 2026 found that 71% of UK SMEs say they haven’t identified a clear use for AI in their business. A cheaper chatbot doesn’t fix that. Knowing what to feed it does.
The gap was never really the model. It’s the context underneath it.
If you’re weighing it up
If you’re looking at a £995 GPT against a full install and you genuinely aren’t sure which your business needs, book a call and we’ll work it out together. No pressure and no catch, even if you just want a second opinion before you spend a penny. Use me as a resource. I’m always happy to help.