"AI chatbot for business" covers two very different things, and conflating them is why so many of these projects disappoint.
Customer-facing chatbots answer your customers on your website. Internal assistants help your team write, summarise and research. Same underlying technology, completely different risk profile and completely different payback.
Customer-facing: what actually works
Triage and routing. Read the incoming message, categorise it, send it to the right queue with a suggested reply. This works well and is low-risk because a human still sends the answer.
Answering documented questions. Opening hours, shipping policy, how to reset a password. If the answer exists in your help centre, a bot can retrieve it.
Out-of-hours holding. Acknowledging a message, gathering details, setting expectations. Better than silence, and honest about what it is.
What does not work: letting a model answer anything about billing, refunds, account status or legal terms without supervision. Those are exactly the conversations where a confident wrong answer costs real money — and the model has no way to know it is wrong.
The single most important design decision
Make it obvious it is a bot, and make the escape hatch to a human visible at all times.
Customers tolerate bots. What they do not tolerate is being trapped by one, or discovering after five minutes that they have been explaining a problem to software. The bots people complain about are almost always the ones that hid what they were.
Internal assistants: the quieter win
Less glamorous, usually higher return, and much lower risk because the audience is your own staff.
- Drafting and rewriting customer emails, for a human to send
- Turning meeting transcripts into action items
- Summarising long research or contracts
- Answering "where is the doc about X"
- Writing and reviewing code
The risk profile is fundamentally different: if the model gets something wrong, an employee catches it before it reaches a customer.
Build vs buy
Buy a support platform's bot if you want customer-facing automation. They handle the integration, escalation and reporting, and that plumbing is most of the work.
Buy a subscription if you want internal assistance. A chat interface with good models is enough for the large majority of teams.
Build custom only when you have proprietary knowledge the model must reason over and enough volume to justify maintaining it. This is a much smaller set of companies than the people selling custom builds suggest.
Grounding, and why bots hallucinate about your business
A general model knows nothing about your company. Ask it about your refund policy and it will invent something reasonable — because inventing something reasonable is what it does when it lacks information.
The fix is grounding: giving the bot your actual documents and instructing it to answer only from them, and to say "I don't know, let me get a human" when they do not cover the question.
Any vendor who cannot explain how their bot is grounded is selling you a hallucination engine pointed at your customers.
What it costs
- Internal assistants: roughly $20 per person per month.
- Customer-facing platforms: typically priced per conversation or per resolution, and this is where costs surprise people. Model the volume before signing.
- Custom builds: engineering time, plus ongoing maintenance as models change. The maintenance is the part that gets underestimated.
Common questions
What is an AI chatbot for business? Two different things: a customer-facing bot on your site, or an internal assistant for your team. They need different tools and carry different risk.
Are AI chatbots worth it for small businesses? Internal assistants, almost always — the payback is fast and the risk is low. Customer-facing bots, only once you have enough repetitive volume to justify the setup.
Will an AI chatbot replace my support team? No, and treating it that way is how these fail. It removes repetitive triage so your team spends time on the cases that need judgement.
How do I stop it giving wrong answers? Ground it in your own documents, restrict it to answering from them, and make it escalate rather than guess. Never let it answer on billing or legal without a human.
Do customers mind talking to a bot? Not if you are upfront and a human is one click away. They mind being trapped or deceived.
How long does it take to set up? An internal assistant, the same day. A grounded customer-facing bot, weeks — most of which is preparing your documentation, not configuring the bot.