- Home
- Capabilities
- Strategy and transactions
- Adversarial Diligence
Adversarial Diligence: AI and technology due diligence for deal teams
We do not audit the thesis. We test whether it holds. A buy-side sprint that tests whether a target's defensibility survives an AI-fluent challenger, and whether the AI upside in the model survives contact with practice.
Traditional diligence tells you what exists. We test what survives.
Code audits, security reviews, market studies and quality-of-earnings work answer important questions about the target as it stands. They do not tell an investment committee whether a funded challenger could cross the moat, whether customers would move if one did, or whether the AI synergies in the model are real.
Defensibility now lives in proprietary data, owned distribution, switching costs and contracts, workflow entrenchment, regulatory position and encoded domain judgment. A business can be highly defensible on replicable technology, and exposed on brilliant technology. That distinction does not show up in a code review.
Three tests, run at the same time
A small operator-led team pressure-tests the thesis from three directions, proving contested points with small proofs of concept rather than asserting them on a slide.
Test the moat
Could it be rebuilt, and would anyone care?
We work out, in engineering detail, how a challenger would rebuild the target's core workflow with current AI tooling, and prove the contested steps. Then the commercial half: switching costs, contracts, relationships and churn history.
Verify the claims
AI claims, measured against reality
The question is not whether something is really AI. It is whether AI is material to the product, its economics, its defensibility and its valuation. We look at where models are used, how they are evaluated, what data they depend on and where deterministic controls remain.
Prove the upside
The synergy case, tried in practice
Every AI lever in the model, from margin lift and throughput to cross-sell and roll-up integration, is scored proven, plausible or pipedream, with proofs of concept for the levers the returns depend on.
Replicable technology is not a weak investment. Exposed is.
The replication work and the commercial read place the target on the moat read. The synergy tests place every lever in the model on the upside read. The quadrant, not the technology alone, tells the committee what to do with the price.
The moat read
Low replicability, high defensibility
Durable moat
Hard to rebuild and customers stay. The multiple holds; the thesis rests on execution, not defense.
High replicability, high defensibility
Replicable product, defensible business
A challenger could build it, but customers would not move. Underwrite the franchise and discount the technology premium.
Low replicability, low defensibility
Technical edge without lock-in
Hard to rebuild today, but nothing holds the customer. A lead, not a moat. Price the erosion.
High replicability, low defensibility
Exposed
Easy to rebuild and easy to leave. Material thesis risk; the price must assume competition arrives.
The upside read
High feasibility, high materiality
Strategic
Belongs in the investment case, marked proven or plausible, with the test that earned the mark.
High feasibility, low materiality
Useful
Real but small. Belongs in the operating plan, not the investment case.
Low feasibility, high materiality
Speculative
Big if true. Stays out of the base case until a proof of concept moves it.
Low feasibility, low materiality
Interesting, immaterial
A good demo, but irrelevant to underwriting. Noted and set aside.
Six workstreams, one investment committee read
Replication blueprint
How a challenger would reproduce the economically important workflow with current tooling, with proofs of concept only where an assumption changes the thesis, and the cost and timeline modeled.
Commercial defensibility read
The moat beyond the stack: switching costs, contract structures, buying behavior, relationship depth, churn and expansion history.
Competitive and roll-up landscape
Named challengers scored on funding, capability, distribution and time to parity, plus who else could enter the category and what a platform thesis could add.
AI architecture and claims validation
Every AI claim scored on where models are used, how they are evaluated, what data they depend on and where deterministic controls remain.
Realities and pitfalls
What current AI cannot do in the target's sector: where reliability, regulation or hybrid engineering protects the incumbent, and where the same limits cap the synergy case.
AI value-creation test
The AI assumptions in the underwriting model, tested rather than asserted. Levers that assume autonomous agents will carry long regulated workflows are flagged early.
Fourteen days from access to the investment committee
Three phases, run in parallel by a small senior team. The clock starts at data room and management access, not at signature.
Thesis brief
By day three: the workflow slice to test, the challenger stack, the AI-claim inventory and the synergy hypotheses worth testing. The deal team confirms focus before the deep work begins.
Evidence file
By day ten: the blueprint stands, with proofs run at contested steps, the commercial read complete and the synergy tests executed.
Decision memo
By day fourteen: a moat scorecard, a synergy ledger, the cost and time to replicate, and a recommendation the committee can absorb quickly, with the evidence behind it.
Adversarial Screen, seven days
Replication blueprint, AI claims validation and the challenger landscape, from public sources and the data room. No target cooperation required.
Adversarial Diligence, fourteen days
All six workstreams from access, with the commercial read scaled to what the process permits.
Partner-led, every time
The people forming the conclusion are the people doing the work. No analyst handoff.
Best fit when
- Software defensibility affects the entry multiple.
- AI is part of management's product story or the margin case.
- A roll-up thesis depends on technology integration.
- The target runs a regulated workflow.
- The deal team needs evidence inside the window, not after it.
Adversarial Diligence does not replace financial, legal, cyber or conventional technical diligence. It works alongside your preferred providers and adds the missing layer.
Questions clients ask
What is Adversarial Diligence?
A buy-side sprint for deal teams and investment committees. It tests whether a target's defensibility survives an AI-fluent challenger, whether its AI claims hold up against evidence, and whether the AI upside in the deal model survives contact with practice.
How long does it take?
The Adversarial Screen takes seven days and works from public sources and the data room. The full Adversarial Diligence takes fourteen days. The clock starts at data room and management access, not at signature.
Does it replace technical, financial or legal diligence?
No. It works alongside your financial, legal, cyber and conventional technical diligence providers and adds the missing layer: whether the technology thesis and the value-creation case hold under current AI capabilities.
Who does the work?
A small, partner-led team of operators who have built, run and sold companies and who build software today. The people forming the conclusion are the people doing the work.
Can you stay after the deal closes?
Yes. We can stop at the investment committee, or stay to lead the value-creation plan we helped test, through our fractional executive and software practices.
Related
Bring us the deal.
Executed, not advised.
The one where technology, AI or automation is carrying part of the multiple or the value-creation case. In a short conversation with a partner, we identify what is worth testing.
Conversations are confidential.