The chatbot your website probably doesn't need

Most B2B service firm websites do not need a chatbot. A site receiving a few hundred visitors a month, selling a considered five-figure service, gains almost nothing from a widget in the corner — and the budget it consumes is usually taken from the thing that actually converts, which is fast, structured follow-up on the enquiries you already get. I build AI systems for a living, and I talk more firms out of chatbots than into them.

Why is the chatbot always the first recommendation?

Because it is the most visible way to buy "AI", and visibility is what sells.

A chatbot sits on the homepage where the MD can see it, demo it to the board, and point to it as evidence of modernisation. Vendors know this, so the chatbot leads the pitch. Meanwhile the automations that genuinely move revenue — enrichment, routing, follow-up, CRM hygiene — are invisible, so they get pitched last or not at all. This inversion, spend allocated by visibility rather than by return, is the pattern I keep returning to in AI Automation for B2B: what actually works: the useful layer of AI in a small firm is mostly plumbing, and plumbing does not demo well.

What does the maths look like for a small B2B site?

Run the numbers before buying anything. A typical UK service firm of 5–50 staff might see somewhere in the hundreds to low thousands of visitors a month, converting to a handful or two of enquiries. Suppose a chatbot lifts enquiries by a genuinely optimistic 10% — that is one or two extra conversations a month, from a tool that commonly costs a meaningful monthly fee plus setup and upkeep.

Now compare the alternative. Response speed to the enquiries you already receive is one of the most lopsided levers in B2B: as an industry rule of thumb, contact rates drop roughly 8x once a response slips past five minutes. A firm answering enquiries in four hours does not need more enquiries intercepted by a bot — it needs the ones it has answered in four minutes. That fix usually costs less than the chatbot and pays back on the first deal.

There is also a buyer-behaviour point. Someone considering a £20,000 engagement does not want to negotiate with a widget. They want evidence of competence, a clear service description, and a low-friction way to book a conversation with a human.

When does a chatbot actually earn its place?

To be fair to the technology, there are narrow cases where it works:

• Genuine traffic volume — thousands of visitors a month, where even small interception percentages produce real numbers.
• Repetitive, answerable questions — pricing tiers, logistics, documentation — where a bot grounded in your own content deflects support load measurably.
• Out-of-hours capture in one specific form: a bot that does one job — takes contact details and books a slot — rather than attempting conversation.
• Support-heavy products, where deflection is a cost line you can measure against.

Note what these have in common: volume, repetition, and a measurable baseline. If you cannot state the metric a chatbot would move, you are buying decoration. A wider tour of which jobs meet that bar is in What AI can actually automate in a service firm.

What should you build instead?

The unglamorous enquiry pipeline. The mechanism: when a visitor submits the form, then an acknowledgement goes out within a minute and the enquiry is written to the CRM with its source recorded. When the enquiry matches your ideal client profile, then it routes straight to the right person with a notification they actually see, and the reply offers a booking link rather than email tennis. When no meeting is booked within a set window, then a follow-up sequence picks the lead up automatically. When nothing happens for 90 days, then the lead enters reactivation rather than oblivion.

Every step is boring, and every step compounds. Firms that install this typically find their "lead generation problem" was substantially a lead handling problem. The tooling behind it is commodity workflow software — compared honestly in n8n vs Make vs Zapier — not a five-figure conversational AI project.

How do you decide, in the end?

Measure the loop before adding parts to it. Enquiries in, response time, booking rate, show rate, close rate — once those numbers exist, the weakest link is usually obvious, and it is almost never "insufficient chat". The discipline of reading your pipeline as a set of loops rather than a wish list is covered in Feedback loops: the physics of your pipeline.

A chatbot is a fine tool for a problem most small B2B firms do not have. Buy the boring pipeline first.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks and what to build first.