01 · Before close
The read
A finding and a rating for each lens, plus value creation hypotheses, written for the investment decision.
Why now
Every target now makes an AI promise. Diligence is where you ask for the proof.
In a deal, AI is upside and exposure at once. It can widen margin and open new revenue inside a business. It can also make what the business sells cheaper to copy, and weaken the position the deal is priced on.
Telling the two apart takes an operator, not a questionnaire. We have run data and AI inside global enterprises and built it hands-on in startups, so we read a target the way we would run it: the data it holds, the systems people use every day, and what it would take to deliver the plan.
The read is written for the investment decision, and built to outlast it. If the deal closes, the same lenses become the roadmap for the value creation plan.
What we test
The lenses, read the way an operator would.
Each lens is read against the thesis: what has to be true for the deal to work, and whether the target can show it.
| Lens | What we look at | What we want to see |
|---|---|---|
| Lens 01 Data position | The data the business creates and holds: its quality, who owns it, and the rights to use it with AI. | Data that is documented, owned, and cleared for the uses the plan depends on. |
| Lens 02 What is real today | Which AI features and tools run in production and are used, and which exist only as demos, pilots, or slides. | Systems in daily use, with quality and adoption measured, not asserted. |
| Lens 03 Value at stake | Where generative AI, agents, and data products could move revenue, margin, and hours, sequenced by dollar impact and risk. | Value tied to named workflows and lines of the P&L, not to a long list of use cases. |
| Lens 04 What it would take | The team, systems, and spend needed to deliver the plan, and what to build, what to buy, and what to leave to a partner. | A plan the current team and systems can carry, or a gap that is sized and costed, with a clear build-vs-buy call. |
| Lens 05 Where AI threatens the model | Which products, services, and workflows a model, or an AI-native competitor, could now deliver cheaper or better. | A position built on data, workflow depth, or customer relationships that a cheaper, AI-native offer cannot copy. |
| Lens 06 Model and vendor dependency | Reliance on one lab, one vendor, or one API, read with a model-agnostic lens: the cost to switch, and what breaks if prices or terms change. | A model-agnostic layer and an evaluation harness, so a switch is a configuration change, not a rebuild. |
| Lens 07 Evaluation and governance | How model quality is tested, who signs off, and how data use, rights, and risk are governed. | Quality tested on the business’s own cases on every change, with a person signing off where it matters. |
What you receive
One memo, written for the investment decision.
Every lens gets a finding, the evidence behind it, and a rating. Then the value creation hypotheses, so the read carries into the hold.
Illustrative template. Not a client document.
Promise & ProofAI due diligence memo
Target: Company name
01 Summary for the investment committee
02 Findings, by lens
- Key
- RRed: a risk to the thesis
- AAmber: a gap with a cost to close
- GGreen: supports the thesis
- Data positionFinding goes here.Rating not filled in.RAG
- What is real todayFinding goes here.Rating not filled in.RAG
- Value at stakeFinding goes here.Rating not filled in.RAG
- What it would takeFinding goes here.Rating not filled in.RAG
- Where AI threatens the modelFinding goes here.Rating not filled in.RAG
- Model and vendor dependencyFinding goes here.Rating not filled in.RAG
- Evaluation and governanceFinding goes here.Rating not filled in.RAG
03 Value creation hypotheses
- RevenueHypothesis goes here.
- MarginHypothesis goes here.
- HoursHypothesis goes here.
04 Roadmap priorities
- Priority 01Priority goes here.
- Priority 02Priority goes here.
- Priority 03Priority goes here.
Illustrative templateNot a client document
Before close, and after
The read carries past close.
02 · After close
The roadmap
A sequenced AI and data roadmap for the hold period, ordered by dollar impact and risk, ready for the value creation plan.
03 · Hold period
The builds
Forward deployed engineers build in a one to six week window, against success criteria set before the first line of code. Every win becomes a playbook the next portfolio company can reuse.
Questions
Questions, answered.
What does the read cover?
It reads the data position and the rights to use it; what AI is real today; the value at stake in revenue, margin, and hours; what it would take to deliver; where AI threatens the model; model and vendor dependency; and evaluation and governance. Each lens gets a finding, the evidence behind it, and a rating, followed by value creation hypotheses.
Do you work with sponsors or with portfolio companies?
Both. We can diagnose across a portfolio with the operating team, and embed forward deployed engineers inside an individual company to build.
Do you favor a particular AI lab or vendor?
No single vendor gets a default; the job decides. We build with whichever frontier model fits the job: Anthropic, OpenAI, Google, Meta Llama, Mistral, or open-weight models. A target’s dependency on any one of them is read with the same model-agnostic lens: the cost to switch, and what breaks if prices or terms change. See every lab, compared.
Which industries do you cover?
Any industry. Our strength is adapting fast to a new business, which is why private equity portfolios, spanning every sector, are a natural fit. Our case studies include media, streaming, sport, and ad tech; they show how the work is done, not where it stops. See industries.
What happens after close?
The read becomes the starting point for the hold. Its value creation hypotheses become a sequenced AI and data roadmap, ready for the value creation plan. Forward deployed engineers then build in a one to six week window, and every win becomes a playbook the next portfolio company can reuse. See AI value creation for private equity.
Next
Bring us a target and the thesis. We will show you what is real, and what it would take.