“Should I automate this?” often gets answered by gut feel. A better decision starts with a measured baseline, an explicit estimate and a test that can prove the estimate wrong.

Step 1: Price what the task costs you now

Take one specific, repeating task. Not “admin”, but “chasing overdue invoices” or “writing quotes”. Then work out its real cost, which is more than the minutes on the clock.

Time. How long does one go take, and how often? Measure several real examples rather than relying on memory.

Your hourly cost. Not your take-home. Use the labour cost relevant to the person doing the task, and record how you arrived at it.

The hidden costs. This is where manual tasks quietly bleed you, and where most ROI sums go wrong by ignoring it:

  • Mistakes. A mispriced quote, a missed reminder, a double booking. What does one error cost when it happens, and how often does it?
  • Delay. Include a cost only where your own records show that delay changed an outcome. Otherwise leave it at zero.
  • The tax on your attention. The task you keep half-remembering to do drains focus even when you are not doing it.

Mark every estimate as an estimate. Unsupported precision is not evidence.

Step 2: Work out what automation actually saves

Now the other side, and be just as honest here or the sum lies to you.

Time saved. Automation rarely takes a task to zero. Measure the automated run, review and correction together, then compare that total with the baseline.

Errors and delays changed. Add a value only after the trial shows a real change and you can explain the calculation. Until then, use zero.

Then subtract the costs of automating. Two of them, and people forget both:

  • Setup time. Record the real setup, testing and training time. Treat it as a one-off cost.
  • The running cost. The tool’s monthly fee, plus any occasional upkeep when something changes.

Step 3: Do the sum

Put the evidence together into a projected monthly net benefit and a payback estimate.

Projected monthly net benefit = (measured time change x hourly cost) + (evidenced outcome change) − (monthly tool cost)

Payback time = setup cost (your setup hours x hourly cost, plus any one-off fee) ÷ monthly saving

The second number is an estimate, not a promise. Usage, maintenance and tool costs change, so revisit it after the trial and at a fixed review date.

Step 4: Decide against your own rule

  • Set the maximum setup cost, acceptable running cost and review date before the trial.
  • Include data risk, adoption and failure handling alongside the financial estimate.
  • Continue only when the measured result clears the threshold you set in advance.

A hypothetical worked example

Suppose a recurring task takes 45 minutes, happens eight times a month and the responsible person’s time is valued at £30 an hour. The current labour cost is £180 a month. If a proposed workflow still needs ten minutes of review each time, the realistic saving is 35 minutes per run, not the full 45.

Now add the actual tool cost and the setup time. Those are inputs, not promises. Replace every number above with your own evidence before you decide.

Where to take this

Pick one repeating task and run it through these four steps. The result is a testable estimate, not a guarantee to act.

If you would like to do this in the room, bring your own task and baseline to The £1,000,000 AI Blueprint in Leigh. See the event page for current booking details.

To make the maths quick, use our automation time-savings calculator. Put in your own baseline, rate and costs, then treat the output as an estimate to verify.