AI pipelines · paid on measured results

Agents do the work.
A verification layer makes it trustworthy.

I'm Kolya Klitovchenko. For thirteen years I've built the systems that run businesses outside of tech — casinos, creator agencies, bank branches, construction, a luxury concierge. Today I rebuild those operations on AI agents, with an expert signing off every unit of work. You pay a share of the difference we measure.

Agent ledgerexample
09:02 intake   new invoice batch · 38 docs
09:03 agent    drafted 36 entries
09:03 risk     2 → human: new vendor, amount > €5k
09:05 expert   accountant signed 36/36
09:06 journal  evidence pack stored
09:20 agent    support: 12 replies drafted

Illustrative. A live feed from my own operation comes next.

13 yrsbuilding operating systems for businesses
6industries
100+staff in one casino install
3 monthsto rebuild a B2B SaaS product alone — still in use
The problem

AI sped up the work. Not the result.

Across 700+ companies, AI coding agents produced 30% more code and zero more finished tasks — the gain disappeared in the review queue. Anthropic rebuilt its own source control for the same reason. The bottleneck is no longer doing the work. It's trusting it.

Your back office is next. That's what I fix.

Sources: Ars Technica· Anthropic

01 · AUTOStandard of done
02 · AIAgent drafts
03 · HUMANExpert signs
04 · HUMANRisk goes to a human
05 · AUTOJournal
Agent · Casino cage reconciliation v1
Does
Reconciles chips, cash and table results at shift close
Can read
Cage system, table tracking, CCTV event log
Never touches
Payouts, player limits, credit
Signs off
Shift cage manager
If unsure
Flags the table, attaches the camera clip, stops

Example. Every agent gets this card, written down before it runs.

Who's behind it

One operator. A fleet of agents.

From sysadmin at a Kyiv hotel to co-founding casino software, leading a platform team, and running my own life on a fleet of agents.

Read the full story →
What I'm building now

I run my own work on what I sell.

An agent operating system I run my own work and life on: Linc, mozok, 15+ open-source CLIs for agents. Every pipeline I sell, I've run on myself first.

See what's live →
Next step

Show me one process that eats your week.

30 minutes. I'll tell you what an agent can carry, who should sign it, and how we'd measure it.

Book a process review