You've been pitched three AI chatbots this year. Here's why they're all the same one.

If you run a business in the UK, you've probably had at least three pitches in the last year from someone offering to build you an AI chatbot, an AI email agent, or an AI-powered lead scraper. The pitches sound different. The websites look different. The products are, more or less, the same.

I want to explain — as someone who builds this stuff for a living — what's actually going on, and what you should look for instead.

What most "AI agencies" are actually selling

Behind almost every AI pitch you've received is one of three products:

1. A rebadged ChatGPT interface trained on your website content. It answers customer questions. Sometimes it hallucinates. It gets removed within six months.
2. A no-code automation built in Zapier or Make.com that emails leads or moves data between apps. Perfectly useful — but nothing an AI-specific consultant is needed to build.
3. A scraper that pulls data from LinkedIn or a competitor's site and generates outreach emails. Frequently against the terms of service of whichever platform it's scraping.

None of these are bad tools. The problem is they're all wrappers around free or cheap APIs, dressed up as bespoke AI systems. That's why the pitches feel interchangeable — because they are.

Why they don't stick in your business

The three most common reasons an "AI pilot" doesn't survive the first six months, based on what I see when I get called in to replace one:

• Nobody redesigned the workflow. The chatbot got bolted onto a broken process. Staff still have to double-check every answer, so the AI adds work instead of removing it.
• It doesn't talk to your other systems. The bot doesn't know about your booking calendar, your inventory, your CRM. So it gives generic answers and customers get frustrated.
• There's no oversight loop. Nobody notices when the AI starts making mistakes, until a customer complains. By then trust is gone.

McKinsey's 2026 State of AI report puts numbers on this: only 6% of organisations using AI actually get material returns from it. The other 94% have adopted something and never seen it pay off.

What you should actually be paying for

If you're going to spend real money on an AI project, three things matter far more than the model or the "prompt":

1. Someone redesigning how your team works around the AI. Not just installing a tool. Actually sitting down and asking: what's the bottleneck, how does it work today, what should it look like after?

2. Real integration with the systems you already use. Your CRM, your accounting tool, your booking calendar. If the AI can't touch them, it's decoration.

3. An oversight layer. Human review for anything the AI is unsure about. Audit logs so you can see exactly what it did. Alerts when quality drops. This is the part that separates a system that lasts three years from one that lasts three months.

What to ask on the next sales call

If you're evaluating anyone selling you AI — including me — three questions cut through the noise:

• "How will you know if it's working after six months?" — if they don't have a concrete answer involving evaluation and metrics, they haven't thought past the pilot.
• "Which of my existing systems will this connect to, and how?" — vague answers here mean vague results.
• "What happens when the AI gets something wrong?" — the answer should include a human review path, not "we'll add a disclaimer".

If the person pitching can't answer those three questions with specifics, they're selling you a wrapper.

This is one of the notes I publish on how AI is actually being adopted in UK businesses. If you're weighing up an AI investment and want a straight opinion, book a 30-minute reality-check call — no pitch, just an honest read on whether it's worth doing.