The short answer: charge a one-time setup fee of $500 to $3,000 plus a monthly retainer of $300 to $1,500, depending on scope. Never sell an AI chatbot as a one-off project. The retainer is the business.
Here's why that structure matters, starting with a cautionary tale.
A guy messaged me last month. He'd just closed his first chatbot client. Super proud. Charged them $300.
One time.
For a bot he now hosts, maintains, updates, and answers support questions about. Forever. I didn't have the heart to tell him he just signed up to be someone's unpaid IT department for the price of a pair of sneakers.
Let's fix the pricing before it fixes you.
The one rule: never sell a bot, sell a subscription
A chatbot is not a website. A website you build once and walk away. A bot runs every day, talks to customers every day, and breaks in new and creative ways every day. It's a living thing. Living things need feeding.
So the structure is always the same:
Setup fee (one-time): covers your onboarding work. Gathering their FAQ, setting the tone, connecting channels, testing.
Monthly retainer (forever): covers hosting, model costs, monitoring, tweaks, and the fact that you're on the hook when something weird happens.
Skip the retainer and you've built a job, not a business. Every month starts at zero and you're hunting new clients just to stand still.
AI chatbot pricing tiers (actual numbers you can steal)
Here's the pricing I see working for client-facing bots right now, after talking to a lot of agency folks running this exact model:
| Tier | Setup fee | Monthly retainer | What's included |
|---|---|---|---|
| Starter | $500 | $300/mo | FAQ + lead capture, one channel |
| Standard | $1,500 | $750/mo | Tool integrations, 2-3 channels |
| Premium | $3,000+ | $1,500+/mo | Multi-bot, custom integrations, white-label, SLAs |
Starter tier: $500 setup + $300/month. A simple FAQ and lead-capture bot for a local business. One channel (web chat or WhatsApp). Answers questions, collects contact info, hands off to a human when stuck. A salon, a gym, a small clinic.
Standard tier: $1,500 setup + $750/month. Bot connected to their actual tools. Books appointments in their calendar, checks order status in Shopify, creates tickets in their helpdesk. Two or three channels. This is the tier most small-to-mid businesses actually need.
Premium tier: $3,000+ setup + $1,500+/month. Multi-bot setups, custom integrations, white-labeled under the client's own domain, SLAs. For clients where the bot touches revenue directly, like sales qualification or e-commerce support at volume.
Notice the spread. The top tier exists partly to make the middle tier look reasonable. Show all three, most clients pick standard. That's not a trick, it's just how people compare prices. No middle option and they anchor on your cheapest number instead.
Why you can charge this (the client's math)
You're not pricing against "what does software cost". You're pricing against what the bot replaces, and what doing nothing costs them.
A part-time receptionist or VA handling messages costs a business $1,500 to $2,500 a month. Covers maybe 25 hours a week. Goes on vacation. Quits.
Your $750/month bot covers 168 hours a week and answers in seconds. When you frame it that way, $750 isn't expensive. It's about $25 a day for a thing that never sleeps. Cheaper than their coffee budget. (I've broken down the full AI-employee-vs-VA comparison if you want the ammunition for that conversation.)
And the doing-nothing cost is real too. A business missing after-hours leads loses actual revenue, every night, silently. If your bot catches even two leads a month that would've gone cold, it usually pays for itself. Put that sentence in your pitch. Loss stings more than gain tempts, and every business owner already knows the feeling of a lead that got away.
Your costs (the part that decides your margin)
This is where agencies quietly win or lose. Revenue is only half the equation.
Per client, per month, your real costs look something like:
- Hosting and infrastructure: $20 to $100 depending how you run it
- Model usage (API costs): $10 to $150 depending on traffic
- Your time: and here's the killer
That last one. If every client bot is a hand-rolled snowflake server you SSH into, your time cost eats everything. Ten clients at 8 hours of babysitting each per month is two full work weeks of unpaid maintenance. Your $750 retainer just became a mediocre hourly rate.
If delivery is repeatable, same setup flow every client, isolated bot per client, hosting handled, your time cost drops to an hour or two per client per month. At that point a 10-client book is roughly $7,500/month of recurring revenue against maybe $1,000 of hard costs and a day of your attention. That's the version of this business that's worth building. I've written about what happens when delivery isn't repeatable and it isn't pretty.
Three pricing mistakes I keep seeing
Charging hourly. The client doesn't want your hours, they want a working bot. Hourly punishes you for being fast. Price the outcome.
Underpricing the first client and never raising. A cheap first client is fine, it buys you a case study. But put a date on it. "Founding client rate, locked for 6 months" gives you a built-in reason to reprice later without an awkward conversation.
No setup fee. Free setup attracts clients with zero commitment who churn in month two. A real setup fee filters for businesses that are serious. Counterintuitive but true: charging more upfront gets you better clients, not fewer.
The short version
Sell a subscription, not a project. Three tiers, push the middle one. Anchor against the cost of a human, not the cost of software. And keep your delivery repeatable so the retainer is margin, not maintenance.
The agencies making real money at this aren't smarter. They just priced it like a service that runs forever, because that's what it is. If you haven't closed your first client yet, the pitch that gets you there is its own skill.
See what repeatable client-bot delivery looks like, free to start.