Clients come with very different requests, understanding, and attitudes toward AI. But our work always starts with an audit.
Why?
Sometimes a person has no idea where to apply AI, but they've heard from someone that AI is phenomenal, and they want it.
Sometimes it's the opposite. The client "knows" what to hand over to AI: "Sales department, three people, we want to speed up proposal preparation."
And sometimes they've already tried and got burned.
Meanwhile, the range of attitudes toward technology is just as wide. Some believe AI is a magic wand. Others think it's incredibly complex. Still others see no value at all.
In short, the starting point is completely diverse.
But there's one point where everyone converges. Nobody knows the real picture of their processes. Even those who come saying "I know exactly what to hand over to AI." Knowing where it hurts and understanding where exactly, how deeply, and in what order to treat it — are different things.
That's why everything starts with an audit.
It varies, but regardless of format — express or detailed analysis — the goal of an audit is always the same: identify 3-5 tasks with maximum implementation impact.
What "impact" means
Time savings, money savings, freeing up people — that's obvious. But the real wow effect is when one qualified employee, armed with AI agents, delivers the work volume that used to take a team and months.
Here are some numbers from real projects:
- Desk market research. Used to take months and 5-6 people. Today orchestrated by 1 person. The rest is done by AI agents: collecting, verifying, matching data, and writing the report. Months turn into hours, and costs are an order of magnitude lower.
- Proposal preparation at an IT company. Used to take 2-3 days and at least 2 people. Today 20 minutes and +30 additional proposals per month without headcount increase. Savings ~200K RUB/month.
- A construction company couldn't keep up with estimating — their main sales tool. 2-5 days per estimate, lots of errors and rework, and lost time. With AI — minutes to develop, 20 minutes to verify. Result: with a well-established lead channel, the number of estimates and sales at standard conversion multiplies.
The most interesting clients for me are those who've already tried AI and got burned, but haven't lost faith in the technology.
Clients lack the competence to define and evaluate the quality of AI implementation services. The trouble is that the same slogans hide specialists with completely different understandings of the problem. Result: a pile of failures.
Things I've seen:
- a solution that works incorrectly, completely ignoring the key function it was built for;
- employee sabotage from those who thought AI would replace them;
- attempts to simply copy a solution that worked well for competitors but "didn't take off" here.
The cause is always that nobody figured out before buying what the business actually needs and where AI will deliver real results — and where it won't.
What the path usually looks like: from first conversation to working agent
- The client comes and explains the task.
- We sit down and define goals.
- Initial audit.
- If we align, we create the spec and contract.
- Then a deep audit: we identify several tasks, choose one priority.
- We build a pilot for it.
- A month of testing.
- We reach stable results and scale to the needed number of users.
Who this works best for
- A company with several back-office domains. Marketing, accounting, management, logistics, sales — at least 2-3 from the list.
- 10-15 people in the back office.
- Revenue — from 20M RUB/year.
- High task density per person.
- Already have experience with burnout.
- Know what bottlenecks and breaking processes are. The company is already trying to stabilize and document processes.
This is the bare minimum. Above — it works. Below — you probably don't need AI yet.