Most businesses don't need a smarter chatbot. They need a smarter business assistant.
When someone asks us for a chatbot, they usually want something quite different from a chatbot. The difference decides everything, including whether you should spend anything at all.
"We need a chatbot." I hear it most weeks. And nearly every time, it isn't quite what the person means.
What they actually want is fewer repetitive emails, faster answers, and customers who feel properly looked after.
A chatbot is one way to get there. Often it's the wrong one.
First, the honest bit: a simple FAQ bot is fine
If all you need is to answer opening hours, delivery information and the same ten questions, you don't need us, and you probably don't need much. An off-the-shelf builder like Chatbase, Tidio or Voiceflow will have a decent FAQ bot on your site in an afternoon. For that job they're genuinely good, and we'll tell you so.
If a thirty-pound tool solves your problem, buy the thirty-pound tool.
Most agencies won't say that, because it doesn't sell a project. But recommending the simple thing when the simple thing is right is exactly how you earn the harder work later.
Where the FAQ bot stops
The trouble starts when the questions stop being generic. "Do you have this in my size?" "What would go with a beef course?" "Where's my order?" "Can you recommend something under fifty pounds?"
To answer those, the assistant has to know your products, read your live stock, understand your policies, and connect to your systems. An FAQ bot can't do that. It was never meant to.
A chatbot answers questions. A business assistant helps customers get things done.
What a business assistant actually is
This is the distinction most people miss, and it's the one that matters:
A generic chatbot answers FAQs. Opening hours, delivery, the common questions, from a fixed script. Off-the-shelf, and fine.
A business assistant understands your products, processes, integrations and customers. It's trained on your knowledge, connected to your systems, and built around how your business actually works.
Those aren't two versions of the same thing. They're different products. One is a widget that answers questions. The other carries out part of your business process.
Most chatbot projects fail before the AI even arrives
Here's the uncomfortable part. The projects that go wrong rarely fail on the technology. They fail earlier, because nobody agreed what "good" actually looks like, or because the business's own knowledge, its prices, policies and real answers, was never pulled together in the first place. Point an assistant at a mess and it will answer confidently from the mess.
The AI is rarely the hard part. Working out how your business actually works is.
So we start with the business, not the model. Someone can copy a prompt or a workflow in an afternoon. What they can't copy is the understanding of how your business actually runs, and that's the thing an assistant needs most.
The geeky bit: what makes an assistant "understand" your business
Three things, working together. First, your own knowledge: your products, prices, policies and past answers, stored so the assistant can search them by meaning (retrieval, usually shortened to RAG) and answer from your actual information, not the open internet. Second, connections: live hooks into your stock system, your CRM, your booking tool, so it can check what's true right now, not what was true when someone last updated a FAQ. Third, a tight system instruction: who it is, what it must never do, and when to hand over to a human.
Knowledge gives it the facts. Connections give it the present. The instruction gives it the rules. Miss any one of the three and you're back to a clever-sounding FAQ bot that guesses.
What it looks like when it's done right
The brief from one wine retailer wasn't "build us a chatbot". It was "customers keep emailing to ask which wine goes with the lamb". Those are completely different problems, and only one of them is a chatbot.
We built the other one: an AI sommelier that recommends bottles with real reasoning, never suggests anything that's out of stock, and learns each customer's taste over time. It isn't answering FAQs. It's doing the job a knowledgeable person on the shop floor would do.
Same idea, different shape, in the AI guide we wove through a whole membership ecosystem. You can see both, and the rest, in our case studies. Every one is an assistant that understands the business, not a bot bolted onto it.
Guardrails: the part nobody asks about
Here's the question that never comes up in the first meeting, and always should: what stops it going wrong?
The moment an assistant can touch your prices, your customers and your systems, "usually right" isn't good enough. A demo that dazzles in a controlled test is a very different thing from something you can safely put in front of a real customer at eleven o'clock on a Saturday night.
So a proper business assistant is built with rules around it: never invent a price or a promise, never share what it shouldn't, stay on topic when someone tries to push it off, and, most importantly, know when to stop and hand over to a person. Get that right and it's an asset. Skip it and one confident wrong answer becomes a refund, a complaint, or a screenshot doing the rounds.
The clever part is easy. The safe part is the job.
That's where we spend the real time, and it's the part an off-the-shelf FAQ bot mostly sidesteps, because it never had access to anything risky in the first place. Which, when a simple bot is all you need, is exactly why a simple bot is the right answer.
So which do you actually need?
Before you commission anything, three questions:
- Are the questions generic, or do they need your live data? Generic means an FAQ bot; live data means an assistant.
- Does it need to do something, or just say something? Saying is a bot; doing is an assistant.
- Would a wrong answer cost you a customer? If yes, you want the controlled, connected version.
If you're answering "generic, say something, low stakes", buy the off-the-shelf bot and move on. If you're answering the other way, a widget will disappoint you, and you'll feel it in the reviews.
The real position
We're not the agency that says everyone needs a custom chatbot. We're the agency that tells you which one you need, sometimes an off-the-shelf bot, sometimes a business assistant, sometimes neither.
Framed honestly, that wins us more of the right work, not less. We'll recommend the simple bot when that's all it needs, and we're the obvious choice the day you realise you need the assistant.
If you're not sure which side of that line you're on, that's exactly the fifteen-minute conversation to have before you spend anything.
Book a quick chat →Related: Most businesses only need three AI tools · See the case studies.
Common questions
What's the difference between a chatbot and an AI business assistant?
A generic chatbot answers FAQs from a fixed script. A business assistant understands your products, processes and customers, connects to your systems (stock, CRM, bookings) and helps customers actually get things done. One's a widget on your site; the other's closer to a member of the team.
Do I need a custom chatbot or an off-the-shelf one?
If the questions are generic and low-stakes, an off-the-shelf FAQ bot from Chatbase, Tidio or Voiceflow is often all you need. If it needs your live data, has to do something rather than just say something, or a wrong answer would cost you a customer, you want a custom business assistant.
Can an off-the-shelf chatbot connect to my systems?
Only in limited ways. Off-the-shelf bots are excellent at answering from a fixed set of questions. Checking live stock, reading a CRM, or reasoning over your product range is where they stop, and where a custom build begins.
Will an AI assistant replace my customer service team?
No. A good assistant handles the repetitive, well-understood questions and hands anything sensitive or unusual to a person, with context. It gives your team time back; it doesn't remove the human where the human matters.