"We already have a chatbot." It is the most common thing we hear when we talk to SME founders about converting website visitors into leads. Sometimes they are right. Sometimes they have a support widget with a canned greeting. Sometimes they have an FAQ tool that handles simple queries. Occasionally they have something genuinely conversational. But in almost every case, what they have is a one-persona tool being asked to do a two-persona job, and the commercial outcomes reflect that.
The distinction between a chatbot and a sales agent is not about complexity or price. It is about intent: who the tool is serving, and what success looks like to it.
The same visitor, two outcomes
Rachel is the operations director at a 25-person accountancy firm. She's researching whether a sales agent could help her firm convert more enquiries from its website. She lands on your site and opens the chat.
With a chatbot: Rachel asks whether the platform integrates with Xero. The bot answers yes. She thanks it and closes the tab. Somewhere in your analytics, a "resolved" conversation is logged. Nothing else happens.
With a sales agent: Rachel asks the same question. The agent answers yes, and then, because it knows she's from professional services and landed on the pricing page, it asks what kind of enquiries the firm is currently struggling to convert. Rachel mentions that prospects often ask detailed questions about compliance scope that the team don't have time to answer immediately. The agent surfaces a relevant case study and asks whether a fifteen-minute call to walk through how that's handled might be useful. Rachel says yes. The booking is made before she leaves the page.
Same visitor. Same opening question. One conversation produced a resolved query; the other produced a booked call.
The four things a sales agent does that a chatbot doesn't
It qualifies. A sales agent works out, in the flow of a normal conversation, whether a visitor is a realistic prospect, what kind of business they run, what they're trying to solve, how close they are to a decision. A chatbot answers what's asked and stops there.
It nudges. Once someone looks like a genuine fit, a sales agent moves them gently toward a commercial next step rather than waving them off with a resolved query. It doesn't push; it guides. The visitor still feels helped rather than sold to.
It routes. When a visitor is well-qualified and ready, a sales agent recognises that and hands them to a conversation that can close, a call, a demo, a human who now has context rather than a cold lead.
It hands off with context. A chatbot produces a transcript. A sales agent produces a warm introduction: this is Rachel, she runs a 25-person accountancy firm, her main pain point is responding to compliance questions at speed, she's seen the Xero integration demo, and she's booked a call for Thursday at two. The human picks it up from there.
Why bolting sales onto a support bot doesn't work
The tools that dominate website chat were built to deflect support tickets at scale. Every design decision flows from that goal: answer the question, close the ticket, move the visitor on satisfied. This is exactly the right architecture for support and exactly the wrong one for sales.
You cannot retrofit the second purpose. A support bot has no model of what the business needs to happen next. It has no concept of qualification, no sense of when to advance a conversation toward a commercial outcome, no mechanism for recognising a high-value visitor and treating them differently from someone who wants a refund. These are not missing features. They are missing intentions.
Adding a more capable language model to the top of a support-shaped tool produces a support-shaped outcome, faster and with better answers, but still commercially inert.
Five questions to tell which one you've actually got
If you want to know whether your current chat is a chatbot or a sales agent, ask these:
- When a visitor leaves the chat, what happened next? Did anything change in your pipeline?
- Does your chat behave differently toward a visitor on your pricing page than one on your help centre? Does it know the difference?
- Can it move a conversation toward a commercial outcome, a booking, a referral to a human, a request for more information, without the visitor explicitly asking for it?
- Does it know when to bring in a person, and does it make that handoff with any useful context attached?
- Does it do anything useful with what it learned in the conversation?
If the honest answer to most of these is no, or you are not sure, you have a chatbot. That is not a failing, it may be exactly what you needed it to be. But it is not what a sales agent is, and it will not produce the outcomes a sales agent produces.
The test is not what it can answer. It is what happens after the answer.