Put in the hours a task takes, what an hour costs and what the system would cost to build and run. It shows the hours returned to the team and how long the automation takes to pay for itself.
How the calculation works
Hours removed a year are the weekly hours on the task, times the share an agent can take on, times the working weeks. The value of that time is the hours removed times the cost of an hour. Payback is the build cost divided by the monthly value left after running costs.
Everything is visible and nothing is sent anywhere. The calculator runs in your browser.
What it leaves out
It counts time only. It leaves out faster turnaround for customers, fewer keying errors and the work the team takes on instead, which in practice are often worth more than the hours. It also leaves out the cost of the person who handles the exceptions the agent cannot, so keep the share realistic.
If the answer is marginal, the honest move is usually not to automate yet. A script, a form or a cleaner process can beat an agent for a small task.
How to fill it in honestly
Measure the hours for a real week rather than estimating them from memory, because people underestimate repetitive work. For the share, count the cases that follow the normal pattern and leave out the unusual ones, since those stay with a person.
For running cost, model usage is often the smaller part. Monitoring, integration upkeep and a weekly review of the agent's work cost more than the tokens for most operational tasks.
Common questions
How do you calculate the ROI of AI automation?
Work out the hours a task takes a week, the share an agent can realistically take on and what an hour of that work costs. Multiply them up to a yearly value, then compare it with the build cost and the monthly running cost. The payback period is the build cost divided by the monthly value left after running costs.
What share of a task can an AI agent realistically take on?
It depends on how standard the inputs are. In the production systems described on this site, the manual work removed ranged from about 60 to 90 percent, with the unusual cases staying with a person. A task with many exceptions will sit lower.
How long does it take for AI automation to pay back?
For a high-volume repetitive task it can be a matter of months, and for a small or irregular task it may never pay back. The calculator shows the payback for your own numbers, and a result over a year or two is a signal to look for a simpler fix first.
What does it cost to run an AI agent each month?
Model usage depends on volume and the model, and you can price it with the LLM API pricing tracker on this site. For most operational agents the larger running costs are monitoring, keeping integrations working and a person reviewing the exceptions.