AI Consulting Services
Most AI consulting ends at a deck. I write the code, wire up the agents, and stay until your team runs the new workflow without me. The proof is my own company: 25 people down to 2, still shipping.
An AI consultant helps a company decide where AI belongs, then puts it into production — strategy, tooling, architecture, and the operating rules around it. Oleg Sotnikov does that work hands-on: he ran the transformation on his own company, AppMaster, taking the team from 25 people to 2 while the platform kept 99.99% uptime for users in 190+ countries and now processes 11B+ tokens a month. Engagements start with a fixed-price Team & AI Audit ($5,000, five business days) and continue as a fractional CTO retainer when you want the rollout done with you.
What an Engagement Covers
Six lines of work. Most companies need three of them, and the audit tells you which three.
AI strategy and roadmap
The questions an AI strategy consultant is hired to answer: where AI pays off in your business and where it doesn't, ranked by payback and risk. You get a sequenced roadmap with owners and dates on it, not a maturity model.
AI adoption for engineering teams
Claude Code and Codex in your repositories, agent pipelines in CI, MCP servers connected to your internal systems. I set the rules, review the first pull requests, and coach the team through the awkward weeks.
Workflow and back-office automation
Support triage, document handling, reporting, onboarding — the repetitive work that quietly consumes headcount. Every workflow ships with a fallback path for the cases the model gets wrong.
LLM architecture, routing and cost control
Model routing, prompt caching, batching, evaluation harnesses, and a token budget per feature. AppMaster runs 11B+ tokens a month, so this is daily operations for me rather than a slide.
AI governance and policy readiness
Written rules for which data may reach which model, vendor assessments, audit logging, and the approval path your legal team will ask about. Enough structure to pass review without stalling the work.
Proof-of-concept first
Nothing gets a rollout budget until a narrow version of it works on your real data. A PoC takes weeks rather than quarters, and it either earns the next phase or dies cheaply.
How the Work Runs
Audit
Five business days at a fixed price. I go through your team structure, workflows, codebase, and LLM spend, then quantify what AI changes here and what it doesn't.
Pilot
One workflow or one team, taken to production on real data. Small enough to finish, big enough to confirm the numbers from the audit.
Rollout and operating rhythm
The pilot pattern spreads to the rest of the company, with weekly reviews, cost dashboards, and the governance rules written down. I stay on as your fractional CTO for as long as that takes.
Why Hire Me
- I ran this transformation on my own company: AppMaster went from 25 people to 2, held 99.99% uptime for users in 190+ countries, and now processes 11B+ tokens a month
- FluxoPay brought AI into a regulated financial-services product without chaos — team costs down, budget freed up for marketing (quote below)
- 25+ years in IT and 7 patents behind the advice — I was building production systems long before there was an AI market to consult on
“I'm really grateful to Oleg for the way he helped us at FluxoPay. One of the biggest wins was reducing our team costs without creating chaos, and that gave me room to move more budget into marketing, which I was very happy about. I was honestly quite worried about introducing AI into the product because we are in financial services, so for me this was a sensitive thing, but Oleg approached it very carefully and in the end my fears were not justified at all. He set things up really well, made the whole process feel much more manageable, and had a real impact on the business.”
Where to Go Next
Narrower pages for the parts of this work people usually arrive looking for.
Frequently Asked Questions
What does an AI consultant actually do?
An AI consultant helps you work out where AI earns its keep, then puts it into production. In my case that covers the strategy and roadmap, the tooling for engineering teams (Claude Code, Codex, agent pipelines, MCP), the LLM architecture underneath, and the governance rules that keep legal comfortable. I write code and review pull requests during the engagement instead of handing over a deck and leaving.
How much does AI consulting cost?
The entry point is the Team & AI Audit: $5,000 fixed, five business days, with a written plan at the end. Ongoing work runs $5,000–10,000 per month as a fractional CTO, or from $3,000 per month for advisory, month-to-month with no lock-in. For context, typical fractional-CTO retainers on the market sit at $8,000–25,000 per month. Full numbers are on the pricing page.
Do you implement, or only advise?
I implement. During an engagement I work inside your repositories, set up the agent pipelines, and review the first pull requests that come out of the new way of working. Advisory-only starts at $3,000 per month if you have a strong team that needs direction rather than hands, but the numbers move fastest when I build alongside your engineers.
Which AI stack and models do you work with?
Anthropic Claude together with Claude Code, OpenAI together with Codex, MCP servers for access to internal tools, and open-weight models where privacy or cost makes them the better fit. Model choice is a routing decision per task rather than a loyalty test: cheap models carry most of the volume and frontier models take the hard calls. I revisit that routing regularly, because the price and capability lines move every few months.
How do you handle a regulated industry?
Carefully, and in writing. We start by mapping which data may leave your perimeter and which may not, then choose models and deployment options that respect that boundary — including open models on your own hardware where the rules require it. Audit logging, human approval on consequential actions, and a documented rollback path arrive with the first pilot rather than after it. That is how it went at FluxoPay, inside financial services.
Find Out Where AI Actually Fits
Five business days, a written plan, and a first pilot you can put in the budget. If AI is the wrong answer for part of your business, the audit says so.
At least $50,000/year in identified savings — payroll, cloud, and LLM bills — or the audit is free.
Related reading
Notes on running AI inside real companies: tooling, cost, and the parts that break.
