Big Companies Are Cancelling Their AI Projects. Small Businesses Are Quietly Winning.
Two things happened in June 2026 that look unrelated and are not.
The first is that the enterprise AI story started falling apart in public. The second is that UK small business confidence fell off a cliff again.
I think the second problem has an answer hiding inside the first one. My name is Iraj, I run InstaWebsite through Codemind LTD, and this is the argument I have been making to every business owner who will sit still long enough to hear it.
The enterprise AI hangover
Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027. The reasons given are escalating costs, unclear business value, and inadequate risk controls.
The same analysis names a problem the industry does not like discussing: agent washing. Vendors rebranding existing chatbots, assistants, and RPA scripts as agents without any real agentic capability. Gartner reckons only around 130 of the thousands of vendors claiming agentic AI are the genuine article.
The rest of the picture is just as bracing. 79% of organisations report difficulty adopting AI — a double digit increase on last year. Only about 29% see significant return from generative AI, and about 23% from agents specifically. Roughly 17% of organisations have actually deployed agents, while more than 60% intend to within two years. That gap is where the cancellations come from.
And when Forbes followed up on the forecast in July, the conclusion was blunt: the failures trace back to management, not model capability. Poor governance. Undefined business value. No operational discipline.
Nobody's agent was too stupid. Their project was too vague.
Meanwhile, in Britain
While that was unfolding, the domestic picture got worse.
The FSB Small Business Index came in at minus 10.8 for the quarter ending June — a fall of 16.3 points on the previous quarter, driven above all by wage costs. Construction sat at minus 20.7. Wholesale and retail at minus 19.5. Accommodation and food at minus 15.9.
ONS data from May showed 66% of businesses with ten or more employees reporting higher staffing costs over the previous three months. Asked how they would handle further rises: 44% said put prices up, 38% said absorb it in margin, and 23% said cut the number of people they employ.
Nearly a quarter of employers are planning to shrink their teams. That is the actual state of play in June 2026.
And on top of it, Making Tax Digital for Income Tax went live in April for sole traders and landlords over £50,000. Quarterly submissions instead of one January return. The threshold drops to £30,000 in April 2027, and £20,000 after that. More admin, on people who already have no time, in the year their costs peaked.
Why the small business is the one with the advantage
Here is the connection.
Every reason enterprise agent projects are being cancelled is a problem of size.
Escalating costs come from projects that sprawl across twelve departments. Unclear business value comes from nobody being able to say which number should move. Inadequate risk controls come from an agent being given the keys to systems no single person fully understands. Poor governance comes from having seventeen stakeholders and no owner.
A plumber in Wigan has none of those problems.
The scope is one workflow. The business value is obvious and countable — missed calls answered, quotes sent same day, invoices chased. The risk surface is small. And governance is one person who knows every part of the operation personally.
The enterprise is trying to boil the ocean with a committee. The small business is trying to stop losing four jobs a week to a voicemail. Guess which project ships.
This is genuinely the first technology shift in my lifetime where being small is the structural advantage rather than the handicap. When websites arrived, the big firms had budgets we could not match. When cloud arrived, same story. This time the constraint is not money. It is clarity of scope, and small businesses have more of that than anyone.
What to do with that advantage
Do not buy "an AI strategy." Nobody needs one. Take the four steps that actually work.
- **Name one workflow that leaks money.** Missed calls after five o'clock. Quotes that take three days. Invoices nobody chases. Pick the one that makes you wince.
- Write down what a good outcome looks like, as a number. "Every enquiry gets a reply in under two minutes." "No invoice goes unchased past fourteen days." If you cannot state it as a number, you are about to start an enterprise-style project that will quietly die.
- Automate that one thing, with you on the approve button. Draft automatically, send deliberately, wherever money or reputation is on the line.
- Measure for a month, then pick the next one. Compounding beats transformation. Every time.
That is not a simplified version of what large companies do. It is the version that works, and they are the ones doing the simplified thing.
Where we fit
InstaWebsite is the entry point, and it is deliberately cheap enough to be a non-decision — £19.99 a month, nothing upfront, no contract. A fully managed Next.js site, a free .co.uk domain, managed hosting, SEO built for AI search as well as Google, and Felix AI capturing leads and booking appointments around the clock. Felix is a narrow agent doing one job well. That is the point.
Promote Your Business is free to list on, human-checked, and built so AI assistants can actually read your listing.
Codemind is for when the workflow is bigger than a website: custom backends, CRM, bookings, payments, and automations mapped to how your business really runs.
We are not selling transformation. We are selling the removal of one specific bottleneck at a time, which is the only thing anybody has ever successfully bought.
While the largest companies in the world spend this year cancelling projects, you could spend it automating four workflows. In twelve months that is not a small gap. That is a different business.