AI process audit
A process audit for AI adoption: a diagnostic that shows where AI delivers measurable impact and how many FTEs it is worth. Two weeks, ₽880,000 excluding VAT: interviews, a register of hypotheses, effect calculated with a triple cross-check, and passports for the priority initiatives.
What the audit costs
- The price is ₽880,000 excluding VAT; ₽1,073,600 with 22% VAT.
- What makes up the price: two specialists × two weeks × 40 hours × ₽5,500/hour = 160 hours of work.
- The price is fixed and does not depend on how many hypotheses we find inside the agreed scope.
- The customer buys a scoped result, not timesheet hours: a list of initiatives with calculations and a launch plan.
- Widening the scope beyond two units is priced separately, before the start rather than after the fact.
What the customer gets
- A management report: where time and money go in the processes, and what of that can be automated.
- A hypothesis register: each with an effect estimate, a timeline and an owner on the customer side.
- Passports for the priority initiatives: problem statement, data sources, success metric, dependencies and risks.
- A data and infrastructure readiness map: where data is sufficient, where it must be collected, what blocks the launch.
- A launch roadmap: data preparation → pilot → measurement → trial operation → rollout.
How the two weeks run
- Before the start we agree the scope: which units and processes are in, and who owns data access.
- Week one — interviews with process owners, collection of facts and measurements, the hypothesis register.
- Week two — effect calculation per hypothesis, passports for the priority initiatives, review with the customer.
- Operations keep running: the diagnosis happens alongside them.
How the effect is calculated
- The unit is an FTE: 1 FTE = 220 working days = 1,760 hours per year.
- Bottom-up — how much of each role's time actually goes into the automatable steps.
- Top-down — the ceiling from unit headcount: how many FTEs exist there at all.
- An applicability check — does the hypothesis clear the threshold of one concentrated FTE: the saving must free a position in one unit, not a minute each across a hundred people.
- Every number carries an evidence status: potential · confirmed by the customer · under pilot verification · confirmed in operation.
What is in and what is out
- In: interviews, fact gathering, hypothesis register, effect calculation, initiative passports, data readiness map, roadmap, presentation of the results to management.
- Out: building the solution, data preparation and cleansing, integrations, the pilot. Those are separate stages with their own price.
- Out: surveying the whole company. The audit runs inside an agreed scope — usually one or two units.
FAQ
What does the AI process audit cost?
₽880,000 excluding VAT; ₽1,073,600 with 22% VAT. The price covers 160 hours of work by two specialists over two weeks and is fixed: it does not grow with the number of hypotheses found inside the agreed scope.
How long does it take?
Two weeks, running alongside the customer's operations. What we need from the customer team is interviews with process owners and access to data — not people assigned full time.
How is the effect measured?
In FTEs: 1 FTE = 220 working days = 1,760 hours per year. Every hypothesis is calculated three ways — bottom-up by role time, top-down by unit headcount, and against the threshold of one concentrated FTE. The gap between the methods is shown rather than hidden: that gap is the search zone.
What is acceptance by potential?
We accept a case on the calculated effect before deployment, not on the fact a year later. An objection counts only as a counter-calculation — so the launch decision rests on a number rather than an impression.
About Sandboxer
Sandboxer is a Russian AI company. 40+ AI projects for 7 enterprise clients in 2 years, >₽1bn confirmed impact. A portal of AI agents for a major petrochemical holding. A team of 20, no venture funding.