T.02Tools · Estimate
Automation savings estimator
A reference page for the estimator: the instrument itself, the formula it runs in prose, the one assumption built into it, and what the result does and does not account for. The buyer's version of the same instrument lives on the AI pages; this page is the method.
Free. No email required. Your numbers are multiplied in your browser and nothing you enter leaves it.
- Cost
- Free, no account
- Runs in
- Your browser, with JavaScript
- First published
- 2026-09-02
- Last revised
- 2026-10-04
The tool
Run the estimator
Pick an industry, the kinds of work that still feel manual, how many people do them and roughly how many hours a week each spends. The result is hours a year across 46 working weeks, with every step of the sum printed. Money appears only if you type an hourly cost, and that figure never enters the shareable link.
people
Count only the people who actually touch this work, not the whole company. Somebody part time still counts as one person.
hours a week each
A rough number is fine. Picture a normal week, not the worst one you had.
your cost, optional
Leave this blank and the answer stays in hours. Fully loaded means pay plus tax, benefits and overhead, divided by hours actually worked. This one is never put in the shareable link and never leaves your browser.
Fill in the numbers first
Your figures
Your own numbers, multiplied out
These are your own numbers, multiplied in your browser. Nothing is sent anywhere, and this is not a projection, an offer, or a promise about your business.
Nothing to work out yet. Pick the areas that still feel manual, then enter how many people work in them and roughly how long it takes them each week.
- People doing this work
- Hours a week each spends on it
- Working weeks a year
- 4652 weeks less about six weeks of holiday, public holidays and sick leave.
- Hours a year going into this work
We have not found a published figure we can stand behind for how much of this kind of work automation actually absorbs. Most of the numbers in circulation turn out to trace back to somebody selling something. So this stops at the hours. The hours are yours and they are right. What comes off them is a question for two weeks inside your business, not for a web page.
- People doing this work, multiplied by the hours a week each of them spends on it, multiplied by 46 working weeks, gives the hours a year going into the work. That is the output.
- If you type an hourly cost, the hours are multiplied by it. If you leave it blank there is no currency on the page at all.
- Hours are rounded to the nearest 10 above 200, and to the nearest 1 below it. Money is rounded to the nearest 100. Nothing else is rounded, so every line can be redone by hand.
The formula
Three numbers, multiplied, and shown
The estimator runs one multiplication and prints it. The prose on the right is the same text the component itself shows under the result, read from the arithmetic module rather than retyped here, and the build checks three worked fixtures against it so the words and the function cannot disagree.
The one assumption built into the sum is 46 working weeks a year: 52 weeks less roughly six for holiday, public holidays and sick leave. It is the only figure on the page you did not type. If it is wrong for your business, redo the last line with your own number.
- 01
People doing this work, multiplied by the hours a week each of them spends on it, multiplied by 46 working weeks, gives the hours a year going into the work. That is the output.
- 02
If you type an hourly cost, the hours are multiplied by it. If you leave it blank there is no currency on the page at all.
- 03
Hours are rounded to the nearest 10 above 200, and to the nearest 1 below it. Money is rounded to the nearest 100. Nothing else is rounded, so every line can be redone by hand.
Printed under every result. These are your own numbers, multiplied in your browser. Nothing is sent anywhere, and this is not a projection, an offer, or a promise about your business.
Assumptions and limits
What the estimate accounts for, and what it leaves out
An estimate like this is useful because it is small. It accounts for exactly three things: how many people, how many hours a week, how many weeks. Everything else about the work is outside the sum, and the list on the right names the parts that matter most. The buyer’s page for the same estimator takes those limits further, into what to do with the number once you have it.
- Accounts for
- Headcount, hours a week per person, and a working year. An optional hourly cost, if you type one, applied to the unrounded hours. Rounding to the nearest 10 hours above 200 and the nearest 100 in money, so every line can be redone by hand.
- Does not account for
- What share of those hours automation would actually absorb. No coefficient is applied, because no published figure survived checking in the verification ledger, and inventing one would turn a fact about your business into a forecast about it.
- Treats as uniform
- Every hour, every person and every week. Ten minutes lost eleven times a day and one clean two hour block count the same here, and a busy week typed in becomes a busy year.
- Says nothing about
- Cost of a build, fee, payback or timeline: no such field exists. Below 120 hours a year the result says plainly that the work is not worth automating yet and offers no next step.
Why the sum stops at hours is documented, figure by figure, in the AI savings statistics verification ledger: of nine widely repeated savings figures, seven could not be published as claimed.
Where else it runs
The same instrument, elsewhere on the site
This page is the tool’s permanent address. It also runs on the AI savings calculator page. That page is written for a buyer deciding what to do with the number: what the arithmetic cannot tell you, why there is no headline savings percentage, and what a next conversation looks like. The instrument there is the same one as here. This page is the method and the citation.
To cite this tool
Praxis Consulting Company, "Automation Savings Estimator", a free hours-a-year estimator run in the browser, revised 2026-10-04, https://praxisconsultingco.com/tools/automation-savings-estimator
Changelog
What changed, newest first
- 2026-10-04Given a reference page of its own under /tools, with the formula, assumptions and limits stated in prose. The arithmetic is unchanged.
- 2026-09-02First published on the AI savings calculator page, degraded by ruling R22 to stop at hours a year with no savings coefficient.
Related
Read next
If the hours look worth a conversation
Bring the sum, with the lines you disagree with marked. The first conversation is free and ends with a plain written list of which parts of the work could be taken over, which could not yet, and what we would build first.
No obligation · a scoping conversation first