What the simulator answers
The payback simulator answers a planning question: if an AI agent took over part of a task your team does today, in which month would the savings overtake what you spend, and what would the net result be after twelve and twenty-four months? You supply the task, the hours and the quote you have, and the tool does the arithmetic.
It suits any repeated task where people spend measurable time, such as answering routine email, preparing quotes, entering orders, reconciling records, scheduling or first-pass document review. The inputs are:
- The number of people who do the task
- Hours per person per week spent on it
- The loaded cost per hour
- The share of the work an agent could take over
- The monthly running cost of the agent, such as licenses, usage and support
- The one-time setup cost
- Months until the agent reaches full effect
You can compare three scenarios. Cautious takes a smaller share of the work, Expected uses the share you entered, and Strong takes a larger share, never above all of the work. The results are hours freed per month, the value of that time per month, the net result per month after running cost, the month when the cumulative net turns positive (the payback), and the net after twelve and after twenty-four months. An animated chart draws the cumulative net over twenty-four months with a payback marker.
You can download a share card as an image, share it, or copy a plain-text summary. The tool counts towards the free Lab Pass, and that progress is kept only in your own browser. Nothing you type is sent anywhere.
How to use it
Pick one task
Choose a single task, not a whole department. A narrow task is easier to measure, and the result is easier to trust.
Measure the time
Count how many people do the task and how many hours each spends per week. Ask the people doing the work, or sample one typical week, rather than guessing from the top.
Set the cost per hour
Use a loaded cost per hour: wages plus employer costs such as payroll taxes, benefits and overhead. Your accountant can tell you the figure you already use for other decisions.
Enter a modest share
Start with the share of the work you are confident an agent could take over, not the best case. The Cautious and Strong scenarios already show the range around it.
Add the quote
Enter the monthly running cost and the one-time setup cost from the quote you have, and the months you expect it to take to reach full effect. Include licenses, usage and support in the running cost.
Read the payback month
Look at the payback month and the twelve and twenty-four month net, then change your weakest input and watch how much the answer moves.
How the numbers work
Each result comes from a short chain of arithmetic. In words:
- Hours freed per month is people, times hours per person per week, times weeks per month, times the share handled, times a ramp factor. The tool uses one fixed average for weeks per month.
- The value of the freed time per month is hours freed times the loaded cost per hour.
- Net per month is that value minus the monthly running cost.
- Cumulative net is minus the setup cost, plus the running sum of the monthly net results.
- Payback is the first month in which the cumulative net is zero or above. The twelve and twenty-four month results are the cumulative net at those points.
The ramp factor reflects the months you entered until the agent reaches full effect. In the early months the agent handles less of the work than at full effect, so early savings are lower, and a longer ramp pushes payback later.
The three scenarios differ in the share of the work the agent takes over: smaller than your entry for Cautious, your entry for Expected, and larger than your entry for Strong, capped at all of the work. Treat them as a range around your assumption, not as three predictions.
How to read payback and the scenarios
Payback is the month when what you have saved overtakes what you have spent. Before that month the cumulative line sits below zero, because the setup cost and the early running costs come first. After it, each month adds its net result.
| Result | What it tells you | What can distort it |
|---|---|---|
| Hours freed per month | Capacity released if the agent takes the share you entered | An optimistic share, or hours that were never really spent on the task |
| Net result per month | The value of that time minus the running cost | A running cost that leaves out oversight time and growing usage |
| Payback month | When the cumulative net first reaches zero | A short ramp, or a setup cost that grows once the work starts |
| Net after twelve and twenty-four months | How the investment looks after one year and after two | Counting freed time as cash that never arrives |
Read the three scenarios as a range. If payback is acceptable in the Cautious scenario, the decision is robust. If it works only in the Strong scenario, the plan depends on an optimistic share and deserves a smaller pilot first. If it does not pay back in any scenario within twenty-four months, the task, the quote or the share needs rethinking.
A long payback is not always bad and a short one is not always good. A quick payback built on a share you cannot defend is weaker than a slower one built on numbers you measured.
What the formula leaves out
The simulator measures time and money. Several real costs and risks sit outside it, and they can change the answer.
The biggest is that freed time is capacity, not cash. It turns into cash only if it replaces overtime or contractor hours, avoids a hire you would otherwise make, or moves people to work that earns money. If it simply disappears into a normal day, the value is real but does not show up in your accounts.
| What happens to the hours | Cash effect | Check before you count it |
|---|---|---|
| Overtime or contractor hours are replaced | A direct saving | Payroll records or contractor invoices for the task |
| A planned hire is avoided | A saving later | Whether the hire was really planned |
| People move to work that earns money | Extra revenue | Whether that work exists and your sales process can absorb it |
| Hours are absorbed into the normal day | Capacity only | Faster replies or fewer errors may still matter, but they are not in the formula |
Other items the formula does not include:
- Oversight: a person reviews the agent's work, handles exceptions and approves risky steps. Add those hours to the running cost or lower the share handled.
- Maintenance: prompts, rules, connections and test cases need updating when your products, prices or policies change.
- Errors and rework: a wrong answer or a bad record can cost more than the time it saved. Quality and risk are not in the formula.
- Change management: training people, agreeing new routines and handling the first weeks of friction all take time.
- Data and access: cleaning up records, setting up service accounts and permissions, and a security review may come before the agent can start.
- Risk and approvals: tasks that touch money, customers or regulated data need approvals and logs.
How to pick a task that is worth automating
A good first task is one where the simulator's inputs are easy to measure and the cost of a mistake is low. Look for these traits:
Rule-bound
The steps and exceptions can be written down, so an agent has something to follow.
Repeated and measurable
It repeats often and you can count the hours, so the simulator inputs are measured and not guessed.
Easy to check
A person can review the output quickly, so mistakes are caught before they reach a customer.
Contained access
The agent needs a few systems with narrow permissions, not access to everything.
Reversible
A wrong result can be corrected without harm, and risky steps wait for approval.
Owned
One named person owns the task, reviews the results and decides when to widen the scope.
The first AI agent picker helps you compare candidate tasks, and the AI readiness audit checks whether your foundations are in place. For help choosing and framing a pilot, see AI consulting, and for building and running the workflow see AI automation. If the task is answering the phone, the missed call calculator is the better starting point.
When this is not a good fit
A payback estimate is useful only when the task and the numbers are real. Be careful, or skip it, in these cases:
- The process is unclear or changes every week. Write it down first, because an agent follows rules and cannot follow a moving target.
- The work is mainly judgment, relationships or physical presence. The time spent there is not what an agent frees.
- The task is rare. A one-off or occasional task seldom repays setup and oversight.
- A person must decide, for example in regulated or high-stakes matters. An agent can prepare the work, but the decision stays with a person, so the share handled should be small.
- You have already decided and want a number to justify it. A calculator can be bent to fit any answer, so use it to find your weakest assumption instead.
- You have a quote but no goal. Decide what success looks like and how it will be measured before you compare prices.
What to prepare for a scope
A short description is enough to start. These make the scope sharper:
- The summary from the simulator, copied into the request form
- The task, who does it and how a typical case goes from start to finish
- Your measured hours and where the measurement came from
- Any quote you have, and what it includes and leaves out
- The tools the task touches and who owns those accounts
- The steps that must always be approved by a person, and who approves them
Request
Send the form with your summary and a few lines about the task. You can also call 773-800-1520, Monday to Friday nine to five and Saturday ten to two Central, in English or Polish.
Short call
We ask about the task, your tools and your approvals, and what a good result would look like.
Written scope
We send a written scope that states the goal, the tools, the approvals and how success is measured, before anything is built.
Pilot before rollout
We run a small pilot, review the logs with you and decide together whether to roll it out.
Ready to start? Use the request form.
Questions answered
How does the AI agent ROI calculator work?
How is the payback month found?
What are the Cautious, Expected and Strong scenarios?
What share of the work should I enter?
What is loaded cost per hour?
Does freed time mean money saved?
What should I include in the monthly running cost?
Does the tool store or send my numbers?
Can I compare two quotes?
Is this a forecast or financial advice?
Turn the estimate into a scope
Send us the task and your numbers. We reply with a written scope that states the goal, the tools, the approvals and how success is measured.