Pre-close · Investor grade

AI Due Diligence

Every deal now comes with an AI narrative. We separate the models that actually run in production from the demos, and tell you what it will cost to keep them running, defend them, and scale them after close.

Typical timeline
Typically 2–4 weeks, compressible to fit a live deal timetable
Best suited to
Private equity, family offices, strategic acquirers, lenders
Commercial model
Fixed fee, agreed before kickoff

What you get out of it

Outcomes

  • A defensible read on whether the AI capability is real, rented, or aspirational

  • The true run-rate cost of models, inference, data pipelines and the people behind them

  • Data rights, licensing and model-risk exposure surfaced before it becomes your problem

  • A post-close plan with the AI work sequenced against the value creation thesis

Scope

What we actually look at

Scope is confirmed with you in writing before the engagement begins. Anything outside it is priced separately rather than absorbed quietly.

01

Capability and substance

  • What is genuinely in production vs. pilot vs. slideware
  • Build vs. buy vs. thin wrapper on a third-party API
  • Model performance claims tested against real evaluation data
  • Key-person concentration in the ML and data engineering team
02

Data and legal exposure

  • Provenance and rights for training and fine-tuning data
  • Customer data usage terms, consent posture and retention
  • Vendor and open-source licence terms that limit commercial use
  • Regulatory surface: EU AI Act, sector rules, state privacy law
03

Cost and scalability

  • Inference, GPU and cloud spend per unit of revenue
  • Contract commitments, credits and reserved capacity
  • Architecture headroom against the growth plan
  • Monitoring, evaluation and incident response maturity

Deliverables

What lands on your desk

Written to be read by the people who have to act on it — an operator, a CFO, an investment committee — not to justify the fee by weight.

  1. 01

    AI due diligence report with red / amber / green findings

  2. 02

    Quantified cost model for the AI stack, current and at plan

  3. 03

    Risk register with owners, severity and remediation cost

  4. 04

    First-100-days technology plan tied to the investment thesis