ROI calculator
What one workflow costs you now, and what an agent could hand back.
Every figure below is yours except one. You supply the volume, the handling time, the hourly cost and the headcount. We supply the automation rate, it is an assumption rather than a study, and you can move it. The formula is printed on the page so you can check the result on paper.
No account, no email required to see the numbers. Nothing you type here is stored unless you ask us to send it to you.
What follows from your inputs
On 100.0 hours a month currently spent on this workflow. Every figure below is a range, because a single number would claim a precision these inputs cannot support.
- Hours returned a month
- 50.6 to 69.0
- Cost returned a month
- $3,643 to $4,968
- Cost returned a year
- $43,718 to $59,616
- Hours a month per person
- 10.1 to 13.8
Against a $5,000 fixed scope pilot, that is a payback of 4.4 to 5.9 weeks on your figures and our rate assumption.
An arithmetic consequence of what you entered, not a forecast and not a result we have measured for you.
The formula
All of it. The first line of each step is the arithmetic, the second is the same arithmetic with your figures in it.
-
Hours spent now
volume a month x handling minutes / 60
100.0 hrs = 200 x 30 min / 60
-
Rate after exceptions
automation rate x (1 - exception rate)
50.6% to 69.0% = 55% to 75% x (1 - 8%)
-
Hours returned
hours spent now x rate after exceptions
50.6 to 69.0 hrs = 100.0 hrs x 50.6% to 69.0%
-
Cost returned
hours returned x loaded hourly cost
$3,643 to $4,968 = 50.6 to 69.0 hrs x $72
-
Payback on a $5,000 pilot
$5,000 / cost returned a month x 52 / 12
4.4 to 5.9 weeks = $5,000 / $3,643 to $4,968 a month x 52 / 12
The exception line is deducted separately and deliberately. Folding it into the automation rate would hide the fact that a share of this work still reaches a person.
Read this before you use the number
What this does not count.
Four things sit outside the arithmetic above. Every one of them makes the real return smaller than the figure on this page, which is why they are here rather than in a footnote at the bottom.
01
Implementation time
Building, testing and running an agent supervised takes weeks of elapsed time and some of your team's attention. None of it appears above.
02
Change management
The work has to be done differently for any of this to hold. People need to trust the output, agree what gets approved and stop keeping the old process running beside the new one.
03
Licence and running costs
Model inference, hosting and any software the workflow touches are ongoing costs. They are small next to the figures above in most cases, but they are not zero.
04
Returned hours are not money yet
An hour handed back is worth something only if it is redeployed: absorbed as extra volume without hiring, or moved to work you would otherwise have paid for. If the hour simply disappears into the day, the saving above never lands.
Read this in plain English
The same numbers written out as two paragraphs, for the person you have to explain this to.
Written from the figures above and nothing else. Changing an input clears it, because a paragraph about numbers that have moved is worse than no paragraph.
Take it with you
The result, your inputs, the assumption and the formula, in one document. We do not keep any of it unless you send it to us.
Email me this
We will send you the summary and Liam or Nick will read it before any call, so you are not repeating yourself. No sequence, no follow-up machine.
A number on a page is not a baseline.
Everything above rests on figures you estimated and a rate we assumed. The only way to know is to measure the workflow before anything is built, which is what week one of a pilot is for.
The worked example is the measurement report a pilot produces in week six, in full, with constructed numbers and a real structure.