Your cohort is full of AI companies. How many of them have built anything?
For accelerators and programmes running an AI-heavy intake, where the partners are technical enough to know they cannot check twenty companies themselves.
Work delivered for
- Singapore Government project
Nearly every application mentions AI, and separating a real system from a prompt behind a landing page takes a conversation that nobody in the programme has time to have twenty times.
The founders who need technical help most are the ones least able to describe what they need, and by demo day it is too late to change anything.
Your investors will ask which companies have defensible technology, and the honest answer for some of them is that nobody has looked.
Bringing in a firm per company does not scale to a cohort, and generic mentorship does not answer the question.
How we work
- IdentifyTwo to three weeks. You get a ranked list of what is worth building, and a longer list of what is not.
- BuildWorking software in your repository, running in your stack, reviewed by your engineers.
- ProveThe same number, measured before and after. If it did not move, the report says so.
DHL
The exception queue nobody could read
How an operation at DHL scale found the gap between what its systems knew and what its people could see in time to act, and what it was worth.
Read the DHL case studyWhat you get
The individual version of this already exists and is free: one technical due diligence memo per fund, inside 48 hours. Start there if you want to see the standard before committing a cohort to anything.
Across a cohort it works differently
- A short memo per company. Whether the codebase matches the pitch, where the team carries risk it has not priced, and what the first six months of engineering will cost. Each one states its own gaps on the first page rather than implying coverage it does not have.
- A comparable read. Same questions, same order, so a partner can put twenty memos side by side rather than twenty different documents.
- A session with the founders, run on their own systems rather than as a talk. Most of them have never been asked how they know a change made the product better, and the ones who cannot answer it find out here rather than in a diligence process.
What the founders get out of it
The questions that separate a real system from a wrapper are the ones their next investor will ask, and most of them are answerable with a fortnight of unglamorous work. Evaluation sets, a cost model that survives real traffic, a note on what happens when the model is wrong. Companies that do that work before the round raise more easily, and the ones that do not tend to find out during diligence.
We publish the full question list openly, with what a strong answer sounds like next to what should worry you. Nothing is behind a form. Read it and decide whether you need us at all.
The honest limitation
A read this size is a read, not an audit. Forty-eight hours per company cannot cover security properly, cannot cover a codebase over roughly two hundred thousand lines with confidence, and cannot assess a model whose quality depends on training data that cannot be inspected. Every memo says so on its first page. Where a company needs more than that, the memo says which one and why.