Judge the product on evidence, not the demo.
Technical due diligence for venture investors is a short, proportionate check of a startup's product and team before the term sheet. In two to three days we test the product with our own inputs, look behind the demo at what is automated and what is done by people, read the code and the infrastructure at a level that matches the round, and interview the founders on how the thing is really built. You get a plain answer on whether the product is what the pitch says, what it costs to run, and what would need to be true for it to scale.
A great demo proves less than it seems
Most demos are honest. Many are also rehearsed on the inputs that work. The questions below are the ones the demo doesn't answer.
Does it work on inputs the founder didn't choose?
How much of the "AI" is a person behind the scenes?
What does each task cost in model and hosting fees?
Is the AI theirs, or a prompt on a public model?
Who wrote the code, and could the team rebuild it?
How many customers use it every week, not just signed up?
of European venture deal value went to AI companies in the first quarter of 2026. Every pitch now has an AI story to test. PitchBook, Q1 2026
of commercial codebases audited had open-source licence conflicts. Startups are no exception. Black Duck OSSRA, 2026
What we check
Proportionate to the round. A seed check is mostly product and founders; a Series B check goes deeper into architecture, cost and team.
The product, tested live
Our own inputs, edge cases and failure modes, timed and recorded. What it does well, where it breaks, and how it fails.
Behind the AI
Which models, prompts and data sit underneath, what is automated and what needs a person, and how quality is measured.
Unit economics of the tech
Model, inference and hosting cost per task or customer, and how that changes with scale and model prices.
Defensibility
What the company owns: data and the rights to it, workflow depth, integrations and anything a well-funded copier would struggle with.
Code and architecture
A walkthrough of the repository and infrastructure: whether it is built to grow, and what would need rebuilding.
The founders as builders
Who built what, how decisions get made, and whether the team can hire and lead engineers at the next stage.
How it runs
Fits between partner meeting and term sheet.
Brief
Hands-on session
Follow-up evidence
Written view and call
How this differs from full technical due diligence
VC technical check
Two to three days, before the term sheet, proportionate to an early-stage company. The question is whether the product and team are real enough to back.
Full technical due diligence
Two to three weeks, usually for control or late-stage deals, with every finding rated and costed for the purchase agreement and the 100-day plan.
For buyouts and growth deals, see technical due diligence and AI due diligence.
One fixed fee per company
A VC technical check is £4k to £8k, fixed, depending on the stage and the depth needed. Funds that use it often can agree a rate for a set number of checks a year.
All prices are fixed and agreed before work starts. The full list is on how we work and pricing.
Led by an operator who has been diligenced, and has bought
Brannan Coady co-founded and ran Netsells, a York software consultancy whose clients included Trainline, Uber and the BBC, through its merger with BGF-backed hedgehog lab. He was Chief Product Officer, COO and then CEO of YourParkingSpace, and sat on the Flowbird Group executive committee through its sale to EasyPark. He has been on both sides of technical due diligence, as the founder being examined and as the investor asking the questions. Specialist security and code-scanning partners join where a target needs them.
About Brannan →Questions from investors
Do VCs do technical due diligence?
Increasingly, yes, especially for AI companies. Many funds rely on partner judgement and a demo at seed, then add a light technical check from Series A, when the product, the team and the cost of running the technology start to drive the valuation. A short external check is often cheaper than one bad investment in a product that was not what it seemed.
How do you evaluate a startup demo?
Use your own inputs, not the founder's; try edge cases and odd requests; time the responses; ask what happens when it fails; ask to see logs, evaluation results and usage data afterwards; and ask what each task costs to run. A demo proves the product can work. Evidence of everyday use proves that it does.
How can you tell if an AI startup is just a wrapper?
Ask what would remain if the underlying model became free and available to everyone. Defensible AI companies own something extra: proprietary data and the rights to use it, deep integration into customer workflows, measured quality the founders can show, or distribution. If the honest answer is a prompt and a user interface, the risk is high.
What does a VC technical check cost?
£4k to £8k, fixed, over two to three days, depending on the stage and depth required. Funds that commission several a year can agree a rate.
Will founders be put off by technical diligence?
Good founders generally welcome it when it is proportionate and run by someone who has built companies. We keep the session to a few hours, explain what we are checking, and share the useful findings with the founders if the investor agrees.