The short version
Chatbot ROI is labour hours saved multiplied by your hourly cost, minus the tool cost. The figure hinges almost entirely on the deflection rate, which is the share of conversations resolved without a human, and vendor-quoted deflection rates are usually best cases on narrow question sets.
The arithmetic, and where it goes wrong
The calculation itself is simple. Conversations, multiplied by the share an agent handles alone, multiplied by the minutes a human would have spent, converted to hours, multiplied by the hourly cost. Subtract what the tool costs.
Every ROI calculator in this category does that. The difference between an honest one and a marketing one is entirely in the default deflection rate, because that single input dominates the result and it is the one nobody can verify for you.
So this calculator ships with no defaults you can be flattered by. You supply the rate, and if you enter something above 80% it says so, because published figures at that level almost always describe a best case on a narrow, well-documented set of questions rather than everything a real customer asks.
What this deliberately does not count
Two things, and leaving them out makes the number smaller and more defensible.
It does not assume you reduce headcount. Most teams reinvest saved time rather than cutting staff, and a calculator that quietly assumes redundancies is measuring something you are not going to do. The labour figure here is the value of time freed, not money removed from payroll.
It also excludes revenue from enquiries you would otherwise have missed. For many businesses, particularly ones where calls arrive after hours, that is the larger number. It is left out because estimating it honestly requires knowing your conversion rate on enquiries you never received, which nobody does. Our missed-call calculator handles that side separately.
The break-even rate is the useful output
More useful than the ROI figure, and rarely shown: what deflection rate would you need just to cover the tool cost.
If the answer is 4%, the decision is easy, because almost any working agent clears that. If the answer is 45%, you are relying on the tool performing near the top of its plausible range before it pays for itself at all, and that is a materially different bet.
This reframes the question from "what might we save" to "how much has to go right", which is the more honest way to look at any tooling decision.
How to get a real number instead
Any calculator is a hypothesis. The only figure that means anything is measured on your own traffic.
Run an agent on your real conversations for a month. Count how many were resolved without a human touching them, and how many needed escalating. That is your actual deflection rate, and it will differ from any published figure in both directions.
This is the argument for a genuinely free tier rather than a demo: you can run the experiment before deciding, and the result is data rather than an estimate. Venbit is free to start with no card for that reason.
Questions about this tool
What deflection rate should I assume?+
We deliberately do not suggest one, because it dominates the result and nobody can verify it for you. Rates above 80% usually describe best cases on narrow question sets rather than everything real customers ask.
Why does it not include reduced headcount?+
Because most teams reinvest freed time rather than cutting staff. A calculator that assumes redundancies is measuring something you probably will not do, so this counts the value of time freed instead.
What is the break-even deflection rate?+
The share of conversations an agent would need to resolve alone just to cover its own cost. It is a more honest framing than a projected saving, because it tells you how much has to go right.
Does it count revenue from enquiries I would have missed?+
No, and that is often the larger figure. Estimating it requires knowing your conversion rate on enquiries you never received. Our missed-call calculator covers that separately.