How much does an AI chatbot cost in 2026?

Short answer, because it’s what you came for: a rule-based bot runs $5,000–$30,000, an AI-powered assistant $75,000–$500,000, and a full enterprise build $200,000 to over $1,000,000.

Now the part that decides whether it was worth it. Most teams file a chatbot under “cost” and try to spend as little as possible. That’s the mistake. Built right, an assistant isn’t a cost line you minimize — it’s a revenue line that also happens to cut cost. The cheapest bot that books nothing is more expensive than a pricier one that sells.

Lead with the revenue, not the savings

Every dollar of support savings is real, and every vendor will quote it. But savings cap out — you can only deflect a ticket once. Revenue doesn’t cap. An assistant that answers the product question and closes the sale, recovers the abandoned cart, books the consult, and qualifies the lead is working both sides of the ledger at once. So size that first: not “what will this save,” but “what will this sell.” If a vendor only talks deflection, they’re pitching you half a business case.

What you’re really paying for over time

The build is one-time. The bill that follows isn’t:

• Maintenance and tuning: roughly $1,000–$5,000 a year, more if it’s custom
• Security patches: $500–$2,500 a month
• Model updates to keep answers current: a few thousand per quarter
• Compliance recertification: $2,000–$10,000 a year in regulated industries
• 24/7 coverage if you staff it: $15,000–$50,000 a year

None of it is in the quote; all of it lands on your P&L. Every time your pricing, products, or policies change, the bot has to be updated or it answers wrong with total confidence. On a DIY build that’s your engineers, forever. On a managed platform it’s bundled into a predictable fee.

The three ways you’ll be charged

Each can be the right answer. The model matters less than what you’re actually buying with it.

The number that maps to your P&L

Two, actually — in this order. First, revenue per conversation: what each handled conversation is worth in bookings, recovered carts, and qualified pipeline. Then cost per resolution: what it costs to close a customer’s need without a human stepping in. A reply is activity; a resolution is the outcome. If a vendor can’t show you both, they’re selling you activity.

What our model is — and what it isn’t

Here’s where the “per resolution” sticker can fool you. Some of the lowest prices are backed by a floor of human agents — a call center with an AI front end. You’re not buying software; you’re renting a contact center by the conversation, at contact-center margins and contact-center consistency.

That isn’t us. RVERE is the AI system itself — one approved AI Assistant across web chat, SMS, social DMs, and a voice line, answering 24/7 with no support floor behind it. When a conversation genuinely needs a person, it escalates to your team, with the context already gathered — your closers and specialists, not a stranger reading a script. We deliver it as a managed platform: a one-time build to stand it up on your channels, then a subscription to keep it running, governed, and improving. You’re buying a system that sells and resolves, not a roster of agents.

Build it, or buy the outcome?

The real question isn’t “how much does a chatbot cost.” It’s who is accountable when it has to keep working six months from now. A custom build is cheaper to start and more expensive to own — you inherit the tuning, the compliance, and the 2 a.m. pages. A managed platform costs more in fees and far less in total cost of ownership, because whoever runs it is on the hook for the outcome.

A “cheap chatbot” is usually the expensive one

A bot that only deflects tickets is a pure cost center, and you’ll always be tempted to spend less on it. The better question: does it also make money — book the appointment, recover the cart, qualify the lead, push the offer? An assistant built to sell, not just to answer, pays for itself instead of trimming a line item. Cost takeout is half the case. Revenue is the other half — and it’s the half most “cheap” chatbots leave on the table.

Industry changes the math

Regulated industries add requirements around storage, integrations, supervision, recordkeeping, encryption, and review. Those are real costs, and exactly where “we’ll review the transcripts later” is not a control. It is cheaper, and far less risky, when governance is part of the runtime: encoded into the system and visible on every important reply, not bolted on after.

How to budget it without getting surprised

• Size the revenue first — bookings, recovered carts, qualified leads — then the savings.
• Add up the recurring (maintenance, security, compliance, staffing) before comparing quotes.
• Ask who resolves the conversation: software, or a hidden human floor you’re paying for.
• Decide build-vs-managed by who’s accountable to the outcome, not by the first invoice.

The cheapest AI chatbot in 2026 is the one that pays for itself — and the ones that pay for themselves are built to sell, not just to answer. Price the outcome, not the software.

Frequently asked questions