What an AI chatbot is really worth
The volume an AI chatbot takes off your team is the return, and it's worth far more than the subscription line costs. Here's how that return is worked out, how much of your queue is deflectable, and how fast it pays for itself.
How much does an AI chatbot cost?
Pricing splits three ways, and which one you are quoted changes what the calculator above is telling you.
Per resolution
You pay for each conversation the bot actually closes. Costs track deflection directly, so the ROI maths is clean — but a bad month of unresolved chats still costs you agent time on top.
Per seat
Priced like the rest of your help desk, per agent per month. Predictable, and it gets cheaper per ticket as volume grows, which is the opposite of per-resolution.
Usage or message based
You pay per message or per API call. Cheapest at low volume and the hardest to forecast, because a single confused customer can burn a lot of messages.
Whichever model you are on, the comparison that decides it is the same: your loaded cost per ticket against the per-ticket cost of the bot. Build-your-own is the option people forget to cost properly — the model API is the cheap part, and the engineering time to keep it grounded in current docs is not.
How chatbot ROI is calculated
Start with cost per ticket, which is your agent's time priced out: the average handle time in hours times the fully-loaded hourly cost (salary plus benefits, tooling, and overhead). Every ticket the chatbot deflects gives that cost back. Multiply the deflected tickets by cost per ticket for the gross return, then subtract what the chatbot costs to run for the net.
return / mo = deflected tickets × $ / ticket − chatbot cost
What deflection rate an AI chatbot hits
Deflection is the share of inbound resolved without a human. That is the whole return. An AI chatbot grounded in your help center routinely takes 30–50% of volume off the queue, with best-in-class teams pushing higher. Here's the whole range; your slider result lands on this same scale.
Low — Most volume still reaches a human. A help center and AI agent have plenty of room to take the repetitive load off your team.
Typical — A solid self-serve baseline. AI grounded in your docs typically pushes this band higher within weeks.
Strong — The band most teams hit with an AI agent answering on top of a well-tended help center.
Best-in-class — Best-in-class. More than half of inbound resolved without a human ever touching it.
Chatbot payback period: how fast it pays for itself
An AI chatbot answers from day one — 24/7, the moment it goes live. The savings start immediately, so payback is usually weeks rather than quarters. The engine here is an AI agent that resolves the repetitive questions instantly, backed by a help center that catches the rest through self-serve. What's left is the work that actually needs a person, plus the agent-hours freed to do it well.
payback = chatbot cost ÷ monthly savings
What a chatbot ROI calculation leaves out
Every number above is a savings estimate, and savings estimates are the flattering half of the picture. Four things move the real figure, and none of them fit in a calculator.
A chat the bot ended is not the same as a question it answered. If the customer gave up and emailed instead, the ticket comes back with a worse mood attached — and most vendor dashboards count it as deflected.
An agent is only as current as the help center behind it. Somebody has to write and maintain those articles, and that time is real even though it never appears on an invoice.
Connecting your docs, testing answers, and setting escalation rules takes days to weeks. Payback starts when the bot goes live, not when you sign.
A wrong answer sent confidently to a frustrated customer costs a refund or a churn, not nine minutes of handle time. This is why a handover rule matters more than a deflection percentage.
The version of this that holds up over a year is an agent grounded in documentation you already maintain, that cites what it answered from, and that hands a customer to a person the moment it is not sure. Deflection you can audit beats a deflection rate you have to take on faith.
Chatbot ROI example: a 2,000-ticket team
Formulas are easier to trust once you have seen them run. This is a mid-sized support team on round numbers — put your own in the calculator above to see where you land.
| Tickets per month | 2,000 |
| Average handle time | 9 minutes |
| Loaded agent cost | $32 / hour |
| Cost per ticket | $4.80 |
| Deflection rate | 40% (800 tickets) |
| Monthly savings | $3,840 |
| Chatbot cost | $800 / month |
| Net monthly gain | $3,040 |
| Payback period | About 6 days |
| Agent-hours returned | 120 hours / month |
