// calculator · savings · AI adoption
Savings calculator: how we estimate the effect of AI
How the savings calculator on tacticgu.ru works: its coefficients, where the range comes from, why there is a turnover cap, and what the estimate leaves out.
in short / answer
The calculator takes the number of employees, the type of business and monthly turnover. It estimates what routine work by office staff costs and assumes AI can take over 15 to 40% of that routine. A cap limits the estimate from above: no more than 15–25% of monthly turnover. The result is a monthly range rounded to two significant digits. It is a rough estimate based on indirect signals, not a promise: only a review of your specific task gives an exact figure.
The home page has a savings calculator. You answer three questions and get a range: how much per month automating routine work with AI can save you. Here we show how it calculates: the formula, the coefficients and the limits. If the figure seems too big or too small to you, this page shows why.
What does the calculator ask?
Three things:
- how many employees you have — a whole number, up to 1000 people; the calculator does not estimate larger companies, and we review such cases separately;
- what the company does — legal services, construction, retail, services, manufacturing or other;
- monthly turnover — about 1, 5 or 10 million ₽, or 20 million ₽ and more (we cautiously treat the last option as 20 million ₽).
How does the formula work?
The calculation runs separately for the lower and the upper bound of the range.
- Monthly cost of routine = employees × share of office staff × 160 working hours × cost per hour × share of routine in their time.
- Effect by headcount = cost of routine × the share that can realistically be automated: 15% for the lower bound and 40% for the upper bound.
- Cap = monthly turnover × 15% for the lower bound and × 25% for the upper bound. It keeps the estimate within a reasonable share of turnover.
- Result = the smaller of the effect by headcount and the cap.
- Both bounds are rounded to two significant digits, so the estimate does not look more precise than it is. The yearly range is the monthly one multiplied by 12 and also rounded.
What coefficients are built in for different industries?
For each type of business the calculator uses three coefficients. These are our reference values, not market statistics, and we will refine them on real projects.
| Type of business | Share of office staff | Full cost per hour | Share of routine |
|---|---|---|---|
| Legal services | 80% | 700 ₽ | 30% |
| Construction | 50% | 600 ₽ | 25% |
| Retail | 60% | 450 ₽ | 30% |
| Services | 70% | 500 ₽ | 25% |
| Manufacturing | 35% | 450 ₽ | 20% |
| Other | 60% | 500 ₽ | 25% |
Full cost per hour is not just salary but everything the company spends on an employee per hour of work.
What does it look like in an example?
A law firm with 10 employees and turnover of about 1 million ₽ per month.
- Cost of routine: 10 × 0.8 × 160 × 700 ₽ × 0.3 = 268,800 ₽ per month.
- Lower bound: 15% of that is 40,320 ₽. The cap of 15% of turnover is 150,000 ₽. We take the smaller one, 40,320 ₽, and round it to 40 thousand ₽.
- Upper bound: 40% is 107,520 ₽. The cap of 25% of turnover is 250,000 ₽. We take the smaller one and round it to 110 thousand ₽.
The calculator will show “40–110 thousand ₽ per month, which is 480 thousand – 1.3 million ₽ per year”.
Now the same firm, but with 100 employees and the same turnover. The cost of routine grows tenfold, and the effect by headcount goes above the cap. So the calculator will show 150–250 thousand ₽ per month: exactly 15% and 25% of turnover. This is how the cap keeps the estimate from drifting away from the scale of the business.
What does the calculator not take into account?
- Your specific task. It does not know which processes you have or which of them are worth automating first. That is why this is an estimate based on indirect signals.
- The cost of implementation and support. Our first solution costs from 50,000 ₽, and support from 10,000 ₽ per month. Compare them with the part of the range that relates to the chosen task, not with the whole amount.
- Time to launch. The effect does not appear on day one: it usually takes 2–3 weeks to the first result, and then the team needs some time to get used to the new routine.
- Money that cannot be reduced to hours. For example, clients who did not leave for a competitor because they got an answer at night. The calculator does not count such effects at all.
Where does the data you enter go?
Nowhere. The calculation runs right in your browser; nothing is sent to a server or saved. If after the calculation you click “Discuss your task on Telegram”, Telegram opens with a ready message that includes your parameters and the range. Whether to send it is up to you.
What should you do with the figure?
Treat it as a guide to whether it is worth looking into at all. If the range is noticeable, the next step is a free task review: together we find the one task where the effect will show up fastest, and calculate it on your own data.
author / CTO, Tactic Guru
Chief technology officer. Builds AI-based systems.