Claude Fable, GPT and the New AI Model Wave: What It Actually Means for a UK Small Business

This month Anthropic launched its Claude 5 family, with a new top-tier model called Claude Fable. OpenAI and Google keep shipping their own upgrades, and every release comes wrapped in bigger claims than the last. If you run a small business, the honest question is not "which model is best". It is "does any of this change what I should do on Monday".

Here is our answer, as engineers who use these tools daily and get asked about them by clients weekly.

The capability jump is real, and mostly irrelevant to you

Each new generation of models is genuinely better at long, complicated work: reading big documents, writing and reviewing code, holding a task together over many steps. Fable and its competitors are the strongest yet at that kind of sustained work.

But most small businesses were not limited by model quality a year ago, and they are not limited by it now. The gap between what the technology can do and what a typical 15-person firm actually uses it for is enormous. If nobody in your business has yet used AI to draft quotes, summarise meetings, tidy up proposals or answer internal questions, upgrading from one model to a newer one changes nothing. The bottleneck is habits and process, not intelligence.

So the first practical takeaway is boring: pick one decent tool, give your team access properly through business accounts rather than personal ones, agree what may and may not be pasted into it, and build the habit. Which brand you pick matters far less than actually doing this.

Where the new models do change things

There are three places where this generation genuinely moves the needle for smaller firms.

First, software is cheaper to build. These models are now very good coding assistants, and that shows up in quotes. Work that took a development team three weeks a couple of years ago often takes a fraction of that now. If an agency is still quoting 2023 prices for 2026 productivity, that margin is going to them, not you. Ask any developer you hire how AI tooling affects their estimates. The good ones will have a straight answer.

Second, AI is now answering your customers' questions. People increasingly ask ChatGPT, Claude or Perplexity to recommend a local accountant, an IT firm or a care agency, instead of scrolling Google. Those assistants read websites directly, and they can only recommend businesses whose sites they can actually read. That has real technical implications, which we cover in a separate article on AI search visibility.

Third, the security picture has worsened. The same tools that draft your proposals also draft criminals' phishing emails, in fluent English, personalised, at scale. The scruffy scam email is disappearing. The defences do not change (MFA, staff training, email authentication), but the excuse that "we'd spot a dodgy email" is now gone.

What to ignore

Ignore anyone selling an "AI transformation strategy" to a 10-person company. Ignore the urge to build a custom chatbot before you have tidied the boring foundations: your data, your website, your security. And ignore the fear of being left behind. This is infrastructure, not a land grab, and the firms that win with it will be the ones that adopted it calmly and kept their basics in order.

What we actually recommend

Our advice to clients this summer is short. Get your team using one mainstream AI tool for everyday writing and admin, with simple rules about client data. Ask your developers and agencies how AI changes their pricing. Make sure your website is readable by AI assistants, because that is where a growing share of recommendations now comes from. And book your staff a refresher on phishing, because the emails have got better.

None of that requires you to understand what makes Fable different from the last model. It requires the same thing good IT has always required: sensible decisions, made deliberately.

If you want a second opinion on where AI genuinely fits your business, and where it does not, book a free 30-minute call. We will tell you honestly, including when the answer is "nowhere yet".