Beyond M&A

    For investors and buyers

    Nobody can confidently price AI in a deal right now.

    Every investor evaluating a SaaS, tech-enabled services or otherwise digitally dependent business now asks the same second question, right after "is this company financially sound?" How real is the AI, and will any of it matter in three years?

    These five short films are the spoken version of our paper, Who The Hell Knows. Pain, aspiration, traps, and what you can actually test for before you sign.

    Five films · about 13 minutes in total

    01 · The pain

    Who the hell knows: nobody can price AI in a deal right now

    3 min watch

    Public SaaS multiples fell from a median 6.2x EV/revenue at the end of 2024 to 3.3x by Q1 2026, while AI-native software still commands double digits. The same ARR, growing at the same rate, is worth a completely different number depending on a label. That is not a knowledge gap in your team. It is the market.

    What to test for

    • · Which side of the valuation split is this deck implicitly asking you to price it on?
    • · Is the AI claim doing valuation work that the revenue quality cannot support?
    • · What would have to be true in three years for today's multiple to still make sense?

    02 · The trap

    Technology is transforming industries, but there is no tech moat

    3 min watch

    This is not a cynical view, it is the industry's own working assumption, argued inside Google itself in the memo "We have no moat, and neither does OpenAI." If the labs building the frontier models do not believe in a technology moat, a Series C company's version of the same claim deserves real scrutiny.

    What to test for

    • · Strip the model out of the story. What is left that a competitor could not buy off the shelf?
    • · Is the advantage in the technology, or in the data, workflow and judgment wrapped around it?
    • · Could a well-funded competitor replicate the core capability in twelve months?

    03 · The aspiration

    AI-native or AI-washed: knowing which is in front of you

    2 min watch

    Two things make AI defensible, and neither of them is the model. It is proprietary data that compounds with use, and an operating model genuinely rebuilt around the technology rather than bolted onto it. Everything else is a label. The goal is to walk out of the room knowing which one you just saw.

    What to test for

    • · Is the data proprietary and compounding, or licensed and available to anyone?
    • · Has the operating model actually changed, or has AI been added to the existing one?
    • · Has anyone assessed the technical team's structure and psychology, not just the product demo?

    04 · How to address it

    Turn the diagnostic into a number

    3 min watch

    Score each dimension one to five and add them up. This is not precision. It is a forcing function against gut feel and against whatever the management deck has already told you to conclude. Above 32 is credibly AI-native. Below 20, the multiple has not been earned by the technology yet.

    What to test for

    • · Score before the management presentation frames your thinking, not after.
    • · Twenty to thirty-one is transitional: real signal and real risk in the same target.
    • · Is there a documented 90-day AI roadmap with measurable milestones, or a vision slide?

    05 · The partner

    The type of buyer changes the diligence you need

    2 min watch

    Diligence has to move earlier and get faster without getting shallower. Only around a quarter of GPs say AI is integrated into their diligence process, and diligence is the one place a meaningful share call it effective. Choosing a partner fit for this belongs before an LOI, not after.

    What to test for

    • · Can your advisor form a view on AI defensibility pre-LOI, in days rather than weeks?
    • · Is a human analyst directing the AI tools, or is the output unsupervised?
    • · Does the work cover the team and operating model, or stop at the architecture diagram?

    Twenty minutes

    Have a live deal where this question is unresolved?

    Bring the target and we will give you a strong, constructive opinion on how real the AI is, before an LOI rather than after.

    Book a conversation

    Score your target

    Turn the diagnostic into a number.

    This is not precision. It is a forcing function against gut feel, and against whatever the management deck has already told you to conclude.

    The diagnostic

    Score the target in front of you

    Eight dimensions, one to five each. Score it before the management presentation frames your thinking. Nothing you enter leaves your browser.

    01 · Proprietary data

    Is the data proprietary and compounding with use, or licensed and available to anyone?

    Public or licensed dataProprietary and compounding

    02 · Operating model

    Has the operating model been rebuilt around the technology, or has AI been bolted onto the old one?

    AI bolted onGenuinely rebuilt

    03 · Team and structure

    Is the technical team structured, incentivised and psychologically set up for continuous change?

    Traditional hierarchySmall, accountable pods

    04 · Architecture

    Can the architecture swap models and vendors, or is it wedded to one provider's roadmap?

    Single-vendor lock-inModel agnostic

    05 · AI roadmap

    Is there a documented 90-day roadmap with measurable milestones, or a vision slide?

    Vision slide onlyMeasured 90-day plan

    06 · Revenue quality

    Is AI revenue contracted and recurring, or pilot income being counted as ARR?

    Pilots counted as ARRContracted and recurring

    07 · Defensibility

    Strip the model out. What is left that a funded competitor could not replicate in twelve months?

    Replicable in monthsHard to replicate

    08 · Governance and risk

    Are data rights, model usage and customer consent documented and defensible?

    UndocumentedDocumented and defensible

    Your score

    0 / 40

    0 of 8 scored

    The paper · August 2026

    Who The Hell Knows

    Twenty pages on why the same growth number prices two completely different ways right now, how to tell an AI-native business from an AI-branded one, and how to choose a diligence partner fit for a market this unresolved.

    • · Part One. The market you're underwriting
    • · Part Two. The targets you evaluate, with a scored diagnostic
    • · Part Three. The partner you choose, before an LOI not after

    We use your email only to deliver the document and occasional insights. Unsubscribe any time.

    Closing thought

    The moat, in the end, is the people directing it.

    The honest posture is neither panic nor denial. It is "who the hell knows, so let's go find out what we can actually test for." If you have a live deal where that question is unresolved, we will give you a strong, constructive opinion on it.

    Start a conversation