H2MHuczyński
Contact

Training

AI in finance

For VC and PE investment teams of 4 to 12. Sourcing, modelling, diligence, memos: 75-minute modules on public data, run by the engineer who writes the code. Prices on this page, the programme document by email.

  • Harpagan
  • porownajdentyste.pl
  • WhiteContext
  • seecontainerpool.ch

Systems running today for a 25-person marketing agency, a clinic marketing agency, a Swiss pool manufacturer and two public comparison sites.

Who this is for

Not for teams looking for a keynote about AI. There is no keynote. Everyone types.

What you leave with

Seven artefacts, not a feeling

  • A shared git repository with everything the team built, with history, owned by you
  • A prompt library for the five tasks your desk repeats
  • A working local setup, or a browser sandbox if IT says no, for every participant
  • A checked LBO skeleton, or a diligence request list and a clause checklist, depending on the track
  • A one-page data rule: what may be pasted, what must be redacted, what stays offline
  • A one-page automation roadmap for the quarter, with named owners
  • A certificate per participant with the title, the dates and the hours, for the L&D file

Prices on the page

Three ways to buy it

Per group, not per seat. The middle one is what most funds pick. Every price below is also in the programme document I send by email, so your L&D or procurement file can quote it.

All prices net; VAT is added where it applies (a business with a valid VAT number in another EU country: reverse charge). Travel and accommodation at cost. Group sessions run on public data only; the 1:1 track may touch live deal data under NDA.

Proof

Teams and people who trained this way

Every card is a repository. Names appear as each person clears them; until then the work is described without them.

Agenda

The modules, 75 minutes each

Not a slide deck. Every module ends with something committed to the repository. The core day is the same for everyone; the track is chosen before day one.

How the hours count

  • 4 modules of 75 minutes
  • 300 minutes of teaching
  • 5 clock hours of teaching plus breaks, 9:00 to 15:00
Core day, 4 modulesEvery participant. No programming background needed.
  1. M1Your desk: Claude, Claude Code, the Excel add-in, plans and permissions75 minA working setup (or the browser sandbox if IT says no) and a written note of which plan may see which data
  2. M2Prompting for analysts: structure, context, repeatability75 minA prompt library for the five tasks your desk repeats, committed to the fund repository
  3. M3Working with financial data: xlsx, csv, pdf; pulling data from company registries and filed financial statements75 minThree years of a listed company's PDF report parsed into a tidy csv, with a check that totals reconcile
  4. M4Case workshop: from data to artefact (model skeleton, investment note or VDR gap list, the group chooses)75 minOne finished artefact per participant in the shared repository
Track A, modelling and sourcing, 5 modulesSecond day of the two-day programme. Prerequisite: the core day and a three-statement model built by hand at least once.
  1. A1Sourcing and pre-screening with public registry APIs and web search75 minA screen of private companies in one industry class, ranked by revenue growth from public statements
  2. A2Excel automation and complex formulas75 minAn iterative revolver and cash sweep formula set, circularity switch included, explained back
  3. A3LBO business logic: tranches, amortisation, returns bridge75 minA debt schedule with three tranches and a returns bridge on a public target
  4. A4Sensitivity and Monte Carlo in the sandbox75 minA 10,000-path Monte Carlo on entry multiple, exit multiple and EBITDA growth, charted, with a written read of the tail
  5. A5Memory, context and the deal folder75 minA CLAUDE.md and folder convention for one live deal, so context is reproducible
Track B, diligence and documentation, 5 modulesAlternative second day. Prerequisite: the core day.
  1. B1Cleaning chaotic financials and standardising statements75 minFive statement layouts from a public filings registry normalised into one schema, with a reconciliation report
  2. B2Contracts: scanning for unfavourable clauses75 minA risk table with citations to clause numbers, from a checklist the fund owns
  3. B3VDR gap detection75 minThe missing-item list and the follow-up email, from a request list and a folder tree
  4. B4Investment memo and deck from the model75 minA three-page IC memo and a ten-slide deck with every number traceable
  5. B5Deal documents: NDA, term sheet, SHA, SPA75 minA redline with a rationale column and a position log the lawyer can read
Track C, portfolio and exit, 2 modulesUsually the remote continuation after the programme.
  1. C1Portfolio monitoring and automatic summaries75 minA scheduled job that reads a monthly reporting pack and writes a one-page variance summary with a flag list
  2. C2Exit: buyer lists and sale materials75 minA buyer long list from public registers and filings, plus a teaser draft

Logistics

How a day runs

Time
9:00 to 15:00, four modules of 75 minutes with breaks. On site at your office or remote.
You bring
Any laptop with a browser. Excel if the modelling track is chosen. A GitHub account, created in the first module if needed.
H2M brings
A sandbox per participant, API tokens for the day, the shared repository, the prompt library, the programme document and the certificates.
Group
4 to 12 people. Above 12 the hands-on part stops working, so it is two groups.
Language
English, and so are the materials.

Data security

What your compliance officer will ask

In writing, before the analyst can book.

Does deal data train the model
Under Anthropic's commercial terms, Claude for Work, Team, Enterprise and the API exclude inputs and outputs from training by default. Consumer Free and Pro plans now default to training unless the user opts out. That is why deal work does not run on a personal Pro account.
Retention
Standard API retention versus a zero-data-retention arrangement, negotiated with Anthropic per organisation. If the fund wants ZDR, that conversation starts before the workshop.
Which plan
Team or Enterprise for the fund, with SSO, audit logs and a DPA. The workshop itself runs on H2M tokens and public data, so no deal data touches a trainer-controlled account.
NDA
Mutual NDA before the scoping call. Group sessions never touch live deal data; the 1:1 track may, and that is what the NDA covers.
Listed targets
If a participant brings a live listed target, the exercise switches to public data. A rule of the room.
Outputs
Every generated memo, redline and valuation is a draft for a human to sign. This sentence is in the programme document.
Recordings
Decided before day one. If recorded, sessions live in the portal, access-controlled, deleted on request.

The trainer

The engineer who writes the code runs the room

Maksym Huczyński · AI engineer, H2M sp. z o.o.

Maksym ships production agent systems, not slideware: a Slack-native AI operator for a 25-person agency, thirteen published distillation experiments including the results that failed, and the public reference collection of production system prompts with 6,208 stars on GitHub.

He writes the training programme, the code and the reply to your brief.

The portal (cohort only)

What makes the fourth week happen

  • Tasks per session, with the artefact to commit
  • The knowledge base: prompts, guardrails, the data rule
  • Recordings, access-controlled, deleted on request
  • A support queue with a stated response time
  • The sandbox, for 90 days

After the training

Within 48 hours of the last module you get three next steps, each with its price. Take any of them, or none: what the team built stays in its repository.

  1. 01

    Agent Care

    €230 / $270 net a month

    We keep the agent or automation your team built running: it keeps working after a model or API change, and the AI bill stays under a cap.

    Plans and SLA
  2. 02

    Changes on a subscription

    from €590 / $690 net a month

    What the team hits next, we ship for it every month: up to 2 small changes a month on Care, one change always in progress on Lane. On top, every plan takes more changes in a relaxed queue, from 4 small ones a month on Care to 8 on Lane: no deadline, done when there is capacity.

    Plans and SLA
  3. 03

    The mentoring quarter

    €2,350 / $2,700 net per quarter, one person

    For the one person who will own the tooling: ten 60-minute sessions on their live work inside a quarter, with review between sessions.

    What the quarter includes

The procurement questions, answered

Everything an L&D or procurement file asks, line by line.

Training requests · answered in writing

The core day is 4 modules of 75 minutes: 300 minutes of teaching, or 5 clock hours, plus breaks, 9:00 to 15:00. The programme document states the same count.

Core day €2,100 / $2,400 net. Two-day programme €3,750 / $4,300 net. A 1:1 session €280 / $320 net. Cohort from €7,500 / $8,600 net. VAT is added where it applies.

Prices are net. VAT is added where it applies; a business with a valid VAT number in another EU country is invoiced under reverse charge.

Leave your email in the form below and I send the programme document within one business day: the title, the hours, the agenda per module, the outcomes, the trainer, the net price and our registration data.

H2M sp. z o.o., VAT ID PL5252680451; the full registry details are in the programme document and on the About page.

Yes, per participant, with the title, dates and hours, for your internal L&D file.

One VAT invoice after the last day, 14-day terms, or 50% on booking for the cohort. Your purchase order number on the invoice if you need it.

Yes, before the brief. Group sessions run on public data only. The 1:1 track may touch your live deal data, and that is exactly what the NDA covers.

Something else? Ask it on the call

Thirty minutes, then an agenda, a date and the price

Two fields. I reply within one business day with a proposed date, the track I would pick for your team and the programme document.

Reply within one business day · no deck

Book a 30-minute call