For family offices, VC funds, angel syndicates and private investors

Wrapper or moat? We can tell the difference.

Many AI companies are a thin layer on someone else's model, and the next model release can erase them. Others own something that compounds. Our founder has built AI companies since 2001. We help you find the second kind, test it before you invest and protect the investment after you do.

Discuss your investment

Across the investment

From the first look to the boardroom.

Engage us for one stage or all four. Each can be agreed on its own.

01
Deal sourcing
Technology companies that fit your thesis, from our network across frontier AI, robotics, space, energy and fintech, screened before they reach your investment committee.
02
Full evaluation
Technology, commercial traction, product market fit and the founding team, brought together in an investment committee memo with a clear recommendation.
03
Monitoring after you invest
Regular review of progress against the plan, the numbers that matter and the technology roadmap, with early warning when something drifts.
04
Board representation
A board or observer seat held on your behalf, with clear reporting back to you and an independent voice when decisions are made.

Six tests

What separates a moat from a wrapper.

These six tests are for AI software. For robotics, energy, space and other deep technology we also assess scientific feasibility, intellectual property, manufacturing readiness, capital intensity, certification and deployment economics.

01
Proprietary data
Does the company own or generate data its competitors cannot buy or scrape, and does that data improve the product with use?
02
Workflow depth
Is the product built into how customers work, with integrations, permissions and processes that are costly to replace, or is it a screen on top of a model?
03
Model independence
Could the company switch model providers without rebuilding, and what happens to the product when the next model release offers the same feature for free?
04
Unit economics after inference costs
Gross margin once model, compute and human review are fully counted, and how that margin moves as usage grows.
05
Distribution
Does the company have a route to customers a competitor cannot copy quickly: channel partnerships, regulatory approvals, embedded integrations or a brand buyers already trust?
06
Path to revenue
Who pays, how much and why now, with evidence from paying customers rather than pilots and letters of intent.

Diligence in practice

Recent screens.

Anonymised. Most companies we look at do not pass. The tests exist to find the ones worth backing, not only to say no.

Medical technology companyConcluded

Strong science, unproven business

The thesis

Wearable safety devices for hospital staff with a cloud platform, presented as the only product of its kind and forecasting rapid growth through distributors.

What we found

The device held up: independent type tests, a letter of intent from a hospital and a leading distributor in its region.
Revenue could not be verified. Sales were never broken down by customer or into one-off and recurring, so concentration and repeat business could not be assessed.
Capital the founders reported investing did not appear on the company's balance sheet.
The AI described was standard data aggregation and reporting, not a source of advantage.
Product market fit was not yet shown: pilots had not turned into orders, and the growth forecast was not built from a customer by customer pipeline.

Recommendation: pass

DisruptIQ engagement

Enterprise AI software companyConcluded

Impressive technology, weak moat

The thesis

Strong revenue growth, recognisable enterprise customers and the promise of AI agents transforming operations.

What we found

Most of the proprietary technology was built on commercial foundation models and open source frameworks, and competitors could increasingly replicate it.
Margins looked attractive until inference costs, implementation effort and customer specific engineering were counted.
Most deployments needed heavy customisation, making the business closer to AI enabled services than a scalable software platform.
The valuation assumed software quality recurring revenue that had not yet been demonstrated.

Recommendation: pass

DisruptIQ engagement

Industrial robotics companyConcluded

A focused first market, and a team that has done it before

The thesis

Humanoid robots for skilled industrial work, starting with a sector facing a structural shortage of qualified labour.

What we found

A founding team with decades of robotics experience in the most demanding environments, including space and deep sea exploration.
A narrow, high value first application with a clear buyer, rather than a general purpose robot in search of a market.
Early partnerships with a major industrial customer and an industry certification body, signalling real demand and a path to qualification.
A robots as a service model that prices against the cost of labour, not the cost of hardware.
Hardware capital intensity and the time to scale remain the key risks, weighed against the size of the opportunity.

Decision: invest

DisruptIQ engagement

Humanoid robotics companyConcluded

A partnership that could not be confirmed

The thesis

A humanoid robotics company presenting a strategic partnership with a major manufacturer as its route to market.

What we found

The partnership, presented as signed, could not be confirmed.
Background checks on the leadership raised issues that had not been disclosed to investors.
Terms and fees were well above market for the stage.

Decision: declined

DisruptIQ engagement

Where to start

Technology Screen

Three to five working days · fixed fee, agreed upfront

You share the deck and data room access. We return a written screen that tells you whether the technology supports the investment case, before you commit to full diligence.

What you receive

The six moat tests scored, with the evidence behind each score
Red flags in the product, architecture, AI exposure and team
The questions to put to management before you go further
A clear recommendation on whether full diligence is worth the cost

Discuss your investment

Also available

Deal sourcing to your thesis · Full evaluation: technology, commercial and product market fit, one to three weeks · Investment committee memo · Monitoring after the investment · Board or observer seat on your behalf

When to call us

You are close to a term sheet on an AI or deep tech company
The valuation depends on technology you cannot assess yourself
You need an independent view of the founders and management team before you commit
The technology is proven, but the path to commercial scale is not
You are co-investing and relying on the lead investor's diligence
A portfolio company's AI story needs independent verification
You need a board or observer seat filled by someone who has built and run technology companies

Decision tools

Test an AI company yourself.

Score a company you are looking at against the six tests. It takes three minutes and shows where to focus your diligence questions.

AI Moat ScorecardTen questions across the six tests: a preliminary screen and the questions to put to management first.3 minutesOpen the tool →

Why our view is useful

Our founder has sat on every side of the table.

Builder. Nikos Patsis founded and exited two technology companies: Voiceweb, Greece's first conversational AI company, and Televida across six Central American countries.
Investor. Chief Operating Officer at Fletcher Asset Management, a private equity firm in New York; Director of Risk Management at Nexus Capital, a global macro hedge fund in Bermuda; and today the DisruptIQ portfolio in frontier technology.
Advisor. Senior advisor on financial and fundraising strategy for more than 35 deep technology projects in the European Innovation Council's Tech to Market programme.

DisruptIQ portfolio

The same diligence we apply to the DisruptIQ portfolio.

The well known names are not the point. Persona AI, Radiant Nuclear, Seldon, Sygaldry, Radium, Reach Power and Odin Space are early stage companies in fields few investors were watching. Each went through the same tests we apply for clients.

Frontier AIAnthropicAI safety and research company that builds the Claude family of models for consumers, developers and enterprises.anthropic.com ↗
AI developer toolsCursorAI code editor from Anysphere that writes, edits and reviews code alongside software engineers.cursor.com ↗
AI product toolsSeldonSimulates how real users will behave, with AI agents built from behavioural data, so product teams can test changes before they ship.seldon.com ↗
AI infrastructureRadiumGPU cloud for AI training and inference, run on its own hardware and tuned for performance, cost and reliability.radium.cloud ↗
Quantum and AISygaldryQuantum accelerated AI servers that pair quantum hardware with classical data centre infrastructure to cut the cost and energy of training and inference.sygaldry.com ↗
RoboticsFigure AIGeneral purpose humanoid robots, controlled by its own AI models, for commercial work and the home.figure.ai ↗
RoboticsPersona AIHumanoid robots for heavy industry, starting with welding and other skilled work in shipyards.persona.ai ↗
SpaceStoke SpaceFully and rapidly reusable rockets, with both stages designed to land and fly again with minimal refurbishment.stokespace.com ↗
SpaceOdin SpaceSensors flown on satellites that detect debris too small to track from the ground, down to a fraction of a millimetre.odin.space ↗
EnergyRadiant NuclearPortable nuclear microreactors that fit in a shipping container, built to replace diesel generators.radiantnuclear.com ↗
EnergyReach PowerWireless power over distance: focused beams that charge sensors, devices and robots without cables.reachpower.com ↗
FintechRevolutGlobal financial app for payments, banking, cards and investing, used by tens of millions of customers.revolut.com ↗