How to calculate the ROI of an AI agent
AI agent ROI tells you whether an agent saves more than it costs. ROI stands for return on investment: what you get back for what you put in. An AI agent is a software helper that reads, thinks and hands back a result. To work out its ROI, you take the value of the time it saves, subtract what it costs, and divide the rest by the cost.
It is the same sum as for a dishwasher
A café owner thinking about a dishwasher asks two questions. How many hours of washing up does it save each month? What does it cost to buy and run? If the saved hours are worth more than the machine, it pays.
An agent is judged the same way. The sum is simple. Honest figures are the hard part.
The formula
ROI = (monthly saving minus monthly cost) divided by monthly cost
- Monthly saving: cases per month, times minutes saved per case, times what a working minute costs you.
- Monthly cost: the software, plus the time someone spends looking after the agent.
A result above 0 means the agent saves more than it costs. A result of 1 means it brings back twice its cost.
Four figures to measure first
- Cases per month. Count them from last month. Do not guess.
- Minutes per case today. Time ten real cases with a watch.
- Minutes per case with the agent. Include the time to read and correct its result. This is never zero.
- Cost of a working hour. Salary plus the employer's extra costs.
A worked example
The numbers below are made up to show the method. Replace every one with your own.
A company receives 400 enquiries a month. Reading and sorting each one takes 4 minutes by hand. With an agent, a person spends 1 minute checking the result. A working hour costs the company 40 euros.
| Line | How it is worked out | Per month |
|---|---|---|
| Time saved | 400 cases × 3 minutes | 20 hours |
| Saving | 20 hours × 40 euros | 800 euros |
| Software | From the invoice | 100 euros |
| Looking after the agent | 3 hours × 40 euros | 120 euros |
| Total cost | 100 + 120 | 220 euros |
| ROI | (800 minus 220) divided by 220 | about 2.6 |
In this example the agent brings back its cost and about 2.6 times as much again.
Is saved time really saved money?
Only if the time is used. Twenty hours saved across five people is four hours each. That can disappear into the week without anyone noticing. Decide beforehand what the freed time is for: more calls, faster replies, a pile of old work that finally gets done. Then check whether that happened.
Three mistakes that make the result look too good
- Measuring the demonstration. Measure a month of real cases, including the awkward ones.
- Forgetting the checking. Someone reads the results. That time is a cost.
- Counting every case. If the agent handles 70 of 100 cases and passes on the rest, only 70 save time.
When a low result is still a yes
An agent may only just pay for itself and still be worth keeping. Perhaps it takes a dull task off someone's desk, or makes results more even. Write that reason down next to the number.
How Worknano handles this
The figures for this sum are already in a Worknano workspace, because agents work on the same records that dashboards and reports read. A log records every agent run, so you can see how often an agent ran. Dashboards show live figures from your workspace, and a click on part of a chart opens the records behind it. Reporting shows how long work takes, how much gets done and how much is waiting. That lets you compare the time before and after an agent joined in.
- Dashboards that are wired to the actual work
- Reporting that answers the questions your teams actually ask
- AI agents and their audit log
Create a free workspace and put your agent's before-and-after numbers on one dashboard.