How to Price an AI Chatbot for Your Clients (4 Real Models)

There are four real ways to charge for an AI chatbot: per conversation, per message, a flat monthly retainer, or an annual contract. Nobody really teaches this part. Each one rewards a different thing, and picking the wrong one can turn your own bot's success into a problem. Here's what each model actually does, and which one to start with.

What are the actual ways to price an AI chatbot?

Four models cover almost every deal we've seen, built, or run ourselves.

• Per conversation (or per resolution). You charge for every conversation the bot fully handles on its own.

• Per message. You charge for every message sent, incoming or outgoing.

• Flat monthly retainer. One number, every month, no matter how much the bot gets used.

• Annual contract. The retainer paid or committed a year at a time, usually at a discount.

What does each pricing model actually reward?

Per conversation rewards volume. The more conversations it closes on its own, the more you make. Sounds great until the client sees the bill jump the exact month the bot started doing its job better.

Per message rewards length, not quality. A bot that solves something in one clean answer earns less than one that drags out over six messages. That's backwards from what you're actually trying to sell.

A flat retainer rewards keeping the thing simple and working. You get paid the same whether it handles 50 conversations or 500, so the job becomes uptime and quality, not volume.

Annual contracts reward commitment. You give a discount for certainty, and the client gets one fixed number to plan a whole year of budget around.

Why does usage based pricing turn against you?

Charge per resolution and the month your bot gets good enough to handle twice the volume, the client's bill doubles too. From where they sit, nothing changed except the invoice. They don't see the training and tuning that got it there, they just see a bigger number for what looks like the same bot. That's the moment a good client starts shopping around, and it was your own success that triggered it.

Charge per message and you get a worse version of the same problem: a shorter, better answer earns less than a longer, worse one. Nobody sets out to build a bot that rambles, but a pricing model that pays by the message quietly points you there anyway.

If your pricing model makes you want your own bot to work worse, that's the wrong model.

What will your actual clients accept?

Coaches, course creators, and community owners already run their whole business on subscriptions: Skool, Kajabi, their own membership fees. A flat monthly number fits how they already think about cost. A bill that moves around, even if it's cheaper some months, reads as risk they didn't sign up for. Most will pick the boring flat number over a technically cheaper but unpredictable one, every time.

How do you actually pick a model for your first clients?

Skip the analysis paralysis. Work through it in this order:

1. Default to a flat setup fee plus a flat monthly retainer. It's the easiest thing to sell and the easiest thing to deliver against.

2. Only go usage based if the client already runs high, predictable volume and specifically asks for it.

3. Never price purely per message. It pays you to make the bot worse.

4. Hold off on annual contracts until the client has run on a monthly retainer for a few months. Prove the build works before you lock in a discount.

---

We don't post one flat number on this page on purpose. Every build is different: what it's trained on, what it connects to, how much a business actually needs it to do. Whether to build one at all matters more up front than the price tag; see Done For You AI Chatbot vs DIY: How to Decide for that call.

If you want to see what a trained agent actually looks like on your own content before you price anything, run our free demo at demo.ajmarketingresults.com.

---

Common questions about pricing an AI chatbot

Should I charge a setup fee separately from the monthly fee?

Yes. Keep them separate. The setup fee covers training the bot on the business's content and connecting it to their tools, real hours that don't repeat. The monthly fee covers keeping it running, updated, and supported. Folding both into one number hides how much of the cost is one-time versus ongoing, and makes it harder to raise the monthly price later without it looking like a hike.

What if a client specifically asks for pay-per-use pricing?

Ask why first. Sometimes it's a real fit, very low and spiky volume where a flat fee would be a waste for them. Often it's just what they know from buying chatbot software, not what actually suits a custom build. If you do agree to it, cap it. Put a ceiling on the bill so a good month for the bot never turns into a shock invoice.

Is per-message pricing ever the right call?

Rarely, for a builder. It's common for chatbot software platforms billing at scale, not for someone hand training an agent for one client. It rewards long conversations over good ones, which is the opposite of what you're actually selling.

Should I offer an annual discount?

Only after the build has run a few months and both sides know it works. An annual deal signed in month one locks in a discount before you've proven the value, and locks the client in before they've seen it work. Earn the annual deal, don't lead with it.

How much do AI chatbots typically cost?

It ranges hard: a couple hundred a month for something narrow and simple, up to real project fees for anything trained on a large amount of content and wired into other tools. There's no honest single number, which is exactly why the model you pick matters more than the number you land on. The model decides how that number moves later.