Your AI pilots,live in 4 weeks.
For operating partners and portfolio CEOs. We map one workflow, delete the waste, build what is left and measure it. Your board gets the before and after.
Now booking December 2026 and January 2027
1 afternoon.A 21-person pharma IT team went from idea to working apps.
- 1 afternoon
- A 21-person pharma IT team went from idea to working apps.
- 10 weeks
- ContactMonkey's first generative AI feature, from zero to production.
Serhii Panchyshyn.Fractional CTO and AI engineer. Venture-backed SaaS since 2019, ex-PwC tax.
Results with names and numbers
- ½ day
A 21-person pharma IT team went from idea to working apps in one afternoon.
- 10 wks
ContactMonkey's first generative AI feature, from zero to production.
- 35%
Lower projected AI cost on that feature, from routing work to the right model.
Background at PwC and Rose Rocket (Y Combinator). Member of the Claude Partner Network.
Where funds get stuck with AI
"We have pilots in three companies. None of them run the business yet."
"The board wants an AI plan. A buyer will ask for proof at exit."
"The company has no technical leader, and a CTO search takes months."
"We bought the licenses. Most of the team still works by hand."
Each quarter of delay is EBITDA you need inside the hold.Most pilots speed up the minutes of workand leave the days of waiting in place.We start with the waiting.
Where the days go
An insurer took 22 days to process an application that needed 17 minutes of work.
A faster AI model speeds up the 17 minutes. The return comes from deleting the wait.
Before: 22 days in the building, 17 minutes of work.
One tool live in 4 weeks
The standard first engagement. One workflow, with one owner on your side. This is everything you get.
Book a 30-minute callThe workflow mapped from your logs
Hours of work and days of waiting on every step, so you see where the money goes.
The tool in production
Inside the systems your team already uses. No new screen and no migration.
A one-page board readout
Before and after numbers your board and a future buyer can check.
A runbook and a handover
Your team runs and changes the tool without us.
Your next five workflows, ranked
With what each one costs you to leave as it is. You decide what comes next.
A person approves every step that moves money
An agent reads each supplier quote and checks it against the contract. A named person approves before anything moves.
Illustration of a procurement app. Three supplier quotes are marked as extracted, a row shows they match the contract, then a card asks the procurement lead to approve. After the approval the card reads approved, tracker updated.
Illustration of the workflow, not a screenshot.
Who this is for
A good fit
- A PE fund or portfolio company with AI pilots that never reached daily use.
- An operating partner or CEO who can name one owner for the workflow.
- Work that runs in Salesforce, NetSuite, Microsoft 365, Slack or similar systems.
- A board that wants the result in numbers.
Not a fit
- You want an AI strategy deck and no build.
- You want a chatbot on your website.
- Nobody can give access to the systems the work runs in.
- You need the whole company rebuilt with AI this quarter.
What you can buy
Prices come after a call, once I know your setup. Each step is fixed in scope, and you can stop after any of them.
For the fund
- Portfolio AI review2 weeks
Where AI pays across your companies, ranked by return, with a plan for the first one.
- AI due diligence2 to 3 weeks
AI risk and margin upside for a company you are buying or selling.
- Hands-on workshop1 day
Your team builds working tools with AI in one session, executives included.
- AI workspace for the firm2 to 4 weeks
Claude for the whole firm, with skills built around how you work and access set by your IT.
For a portfolio company
- First tool live4 weeks
One workflow in production, measured before and after.
- Agentic workflowsafter the first tool
Procurement, fundraising and financial models inside Salesforce, NetSuite or Microsoft 365. No new screen and no migration.
- Fractional CTOmonthly
Roadmap, architecture, vendors and hiring, plus the builds.
- Custom software and SaaSscoped after a review
Including products your company sells to its own customers.
Recent work
Client names stay private until they agree. References are available under NDA.
| Client | What we built | Result |
|---|---|---|
| Public pharma company, IT team of 21 | Hands-on AI workshop | Working apps in one afternoon |
| ContactMonkey, $55M Series A SaaS | First generative AI feature | Live in 10 weeks, 35% lower projected AI cost |
| Rose Rocket, Y Combinator logistics SaaS | Operations workflow automation | 10 hours a week back for the ops team |
| Private equity firm | Claude workspace with the firm's brand and document standards | On-brand, investor-ready documents from a prompt. Signed off in writing in under two weeks |
First, map the real process.
From your system logs and interviews, with the hours of work and the days of waiting on each step.
Then delete the waste.
Steps that exist only for rework or handoffs go before anything gets automated.
Build what is left.
Plain code first, agents where judgment pays, and a person on every decision that moves money.
Measure it again.
Tested on cases with known answers, run next to the old way, and measured again at 30 days.
I own your result, first call to handover
I started in tax compliance at PwC. Partners there asked one question about every file: how do you know this is right? Then I taught myself to code and spent almost six years at Rose Rocket, a Y Combinator company that has raised $69M. Now I work with private equity funds and their portfolio companies. Every tool I build ships with a test report that answers the PwC question.
- Started
- PwC tax compliance
- Built
- Rose Rocket (Y Combinator)
- Member
- Claude Partner Network
"He gets results stunningly quickly."
Questions funds ask first
Tap a question to open it. Anything else, ask on the call.
Why not a big consulting firm?
Their first deliverable is often a strategy deck. Here the first deliverable is a working tool in 4 weeks, with its before and after number.
Why not hire an engineer?
A senior AI hire takes months to find. I can start within weeks, and your team learns to run what we build.
What is an agentic workflow?
Software agents do the reading and typing in a process. A person approves every step that moves money or reaches a customer.
Who owns the code?
You do. It lives in your accounts from the first day.
Where does our data live?
In your accounts. Your admins grant named access and remove it when the work ends. Nothing sends, signs or pays without a person.
Can you move fast on a large build?
Yes. The team scales to your timeline, and you keep one point of contact and one owner for the result.
What do you need from our team?
One owner for the workflow, access to the systems it touches, and about an hour a week of their time.
The company already has a CTO. Where do you fit?
I work for the CTO. They keep the roadmap, and I add the AI build capacity their team lacks today.
How does it end?
With a runbook and a handover to your team. You can stop after any step.
Pick one pilot.Live in 4 weeks.
Book 30 minutes and bring one workflow that is stuck. We work out what the waiting costs you, and you leave with a yes or no on fit. If it is a yes, a written proposal follows.
Now booking December 2026 and January 2027