Khumo Automation · Munich

The deal intelligence, automated.

We build AI for high-stakes deal work — on both sides of the table. Khumo turns confidential documents into IC-ready analysis and professional pitchbooks, with a banker’s rigor, page-level citations, and zero data retention.

~4 hrs → ~5 min
First-pass screen of a CIM — without the analyst hours.
Zero retention
Your documents are never stored and never train a model.
Best models
Model-agnostic routing across Claude, GPT and Gemini.
Built in collaboration with leading dealmakers · Member of
01 · The problem

Early-stage deal review is still done by hand.

Small investment teams receive far more deal documents than they can realistically review in depth. Every CIM forces the same trade-off:

  • Burn scarce analyst hours on weak deals, or risk missing a material issue on first-pass screening.
  • Review is manual and inconsistent — quality varies by analyst, and it’s slow.
  • Lean teams still have to beat better-resourced competitors to the right deals.
The question every deal must answer
“Am I the logical buyer — and can I achieve outsized returns?”
Today, answering it means hours of manual document review per opportunity. We make it minutes.
02 · What we build

Two tools. Both sides of the table.

Both live at app.khumo.ai — open in your browser, nothing to install.

Buy-side · for PE & family offices

Khumo Screen

Drop in a CIM. Specialist agents run triage, deterministic financial extraction, market and valuation analysis, a fit check against your own criteria and a red-flags sweep — then emit a screening memo, an IC one-pager and a deck, every claim cited to its source page.

  • Screening memo (DOCX), IC one-pager & deck (PPTX) — citation-linked
  • 30 KPIs through a deterministic engine — extracted, not guessed
  • Zero Data Retention end-to-end (CIMs under NDA)
Sell-side · for owners & advisors

Instant Pitchbook

Selling your company? Skip the €2,000 banker fee. Upload your documents and Khumo builds a professional sell-side pitchbook in minutes — filled from your data, with Khumo’s PE & IB expertise. Anything we can’t ground is clearly marked to complete.

  • No login — instant, online, a fraction of the cost
  • Grounded in your documents — never fabricated; gaps marked “to complete”
  • Fully editable PPTX · processed in-memory & deleted after download
On the roadmap Coming soon

Khumo Atlas — the data layer beneath the deals.

A per-firm context layer that turns a messy document corpus into an AI-ready, provenance-stamped knowledge map — so your team and your tools know what each file is, what still matters, and what to ignore.

Get early access →
03 · Why Khumo

We don’t compete with the models. We make them trustworthy.

Khumo runs on Claude, GPT and Gemini — and wraps them in the system that makes them reliable for deal work: specialist agents, deterministic math, adversarial verification and page-level citations. Fluency isn’t analytical reliability, and in high-stakes deals that difference is the whole game. It’s the same engine behind both products.

agent ▸ view_triage_map()
agent ▸ read_financials()
agent ▸ run_calculation()

Specialist agents, not one prompt

Screen runs triage → financials → market → valuation → fit → red flags. The Pitchbook runs explorer agents that mine your documents and a composer that builds the deck. Each agent has one job.

48.2 35.8 = 12.4EBITDA

Deterministic financials

KPIs are computed by a calculator engine — extracted, not guessed. The model reads the numbers; the math is deterministic. No hallucinated EBITDA.

Revenue48.2
COGS35.8
EBITDA12.4
48.235.812.4

It won’t invent numbers

Financials are extracted from the page, not computed by a model that might mis-map costs. The Pitchbook leaves a field blank rather than fabricate one. A missing number beats a confident wrong one.

62% recurring p.42

Cited to the source page

Every claim links back to the exact page it came from — provenance you can defend in front of an investment committee.

AnthropicOpenAIGeminiGrokMistralMeta AnthropicOpenAIGeminiGrokMistralMeta

Model-agnostic by design

We route across Claude, GPT and Gemini and pick the best per task. You ride the frontier and hedge the model war — better models only make us stronger.

Confidential by architecture

Zero Data Retention end-to-end. Your documents are never stored and never train a model — built for CIMs under NDA.

Why the last 20% is the only part that matters

Ask a frontier model for a casino’s gross margin and it will confidently mis-map costs and hand you 91%. Khumo doesn’t let the model do the math — it extracts the reported figures and computes KPIs deterministically, so the confident-but-wrong number never reaches your IC. Closing that gap is the whole reason Khumo exists.

Model-agnostic routing Deterministic financials Extract, don’t compute Source-linked citations Configured per firm — no fine-tuning Zero data retention
04 · Pricing

Usage-based. No seats, no lock-in.

Pay for what you run. Screen by the deal or by the month; build pitchbooks one at a time.

Khumo Screen · buy-side

Screen by the deal, or by the month.

Pay-as-you-go
Pay only for the deals you screen. Full IC-ready output every time.
€49 / screen
Subscription for teams
Usage-based — ≈ 57 screens included each month.
€500 / mo
Instant Pitchbook · sell-side

One deck, one price.

€39 / pitchbook
Pay per download. A full sell-side deck, fully editable in PowerPoint and yours to keep — no login, no subscription.

Prices exclude VAT where applicable. Building pitchbooks at volume? Talk to us.

05 · The team

Builders who know the buyer.

Frederick Erleigh
Frederick Erleigh
Founder & CEO
AI meets client needs — product & financial services.
ProductLMU
Kalman Mahlich
Kalman Mahlich
Co-Founder & CTO
ML & scalable systems; LLM integration & deployment.
ML systemsUT Austin · TUM
Julian Seifinger
Julian Seifinger
Co-Founder & CSO
B2B sales, venture building & partnerships.
GTMLMU
Christopher Erleigh
Christopher Erleigh
Founding Engineer
Front-line engineering; genetics & programming.
EngineeringUCT
“Not only do they bring strong technical expertise to the table, but they also understand how the private equity industry works. Khumo is constantly evolving, addressing real pain points, and in our view has great potential.”
Alexander Rees — Investment Associate, Pinova Capital

See it on your own deal.

Open the app to run a screen or build a pitchbook — or reach out for a walkthrough. We’re a team of dealmakers and engineers in Munich, and we answer fast.