RC HEV Propagation Compliance Deal Export AI Ready Culture 9AI Disruption

Enterprise Reconstruction · Module 9

You are not just buying the enterprise.
You are buying its exposure.

AI Readiness asks whether this enterprise can adopt AI. This asks a colder question: how much of what you are paying for can a general-purpose model already do — or will be able to do within the hold period? That number does not show up in the P&L you were handed. It shows up eighteen months after close, in the multiple you can no longer get.

Move each parameter to reflect the target as it actually operates today — not as its pitch deck describes it.

If the value delivered to the customer is mostly a general-purpose model with a thin layer on top, the moat is thinner than the multiple assumes.

Core value delivered by tasks a frontier model can already perform0 = deeply proprietary workflow · 100 = thin wrapper on a general model 35%

Proprietary, hard-to-replicate data and validated expertise are the actual defensibility. Public or easily scraped data isn't.

Strength of proprietary data / expert-validated content advantage0 = public data, easily replicated · 100 = decades of proprietary, licensed, or expert-validated content 50

The newer risk isn't a model answering one question. It's an agent chaining several steps end-to-end and removing the product entirely, not just one feature of it.

Risk that an agentic system replaces the full customer workflow, not one task within it0 = workflow requires irreplaceable human judgment · 100 = fully chainable by an agent today 40

A data moat with no one left to maintain it degrades quickly. This mirrors the Knowledge Concentration Index in Reconstruction Cost — the same fragility, aimed at AI-specific expertise.

Share of AI/ML-critical expertise concentrated in employees likely to leave post-close0 = distributed, documented, replaceable · 100 = two or three people, undocumented 45

Every exposure has a clock. PwC's 2026 guidance to dealmakers is to price AI's impact over a three-to-five year window, not the trailing twelve months on the CIM.

Years before this exposure is likely to materially affect revenue or retention1 = already happening · 7 = distant, low urgency 4 yrs
Move the parameters and calculate to see the exposure score.

AI Disruption Exposure Score

Time-adjusted urgency

Model dependency

Data moat (inverted)

Agentic substitution risk

Talent concentration

Precedent · Not a hypothetical

Thomson Reuters bought its own disruption first.

In 2023, Thomson Reuters paid $650 million in cash for Casetext — a 104-person legal AI company that, by the acquirer's own analysts' account, was not expected to move the revenue needle in the near term. The deal was projected to be margin-dilutive through the following year. Thomson Reuters bought it anyway.

The reason maps directly onto this tool. Thomson Reuters' core franchise — Westlaw, legal research — was itself exposed to model-dependency risk: a general-purpose model could increasingly do what a paid legal research seat did. Casetext's product, CoCounsel, had early access to GPT-4 and had already proven the workflow could be automated end-to-end. Rather than let a competitor or a startup own that substitution, Thomson Reuters bought the exposure and put it inside its own moat — its proprietary, expert-validated content and its existing customer base of over 10,000 firms.

$650M
Cash price, 104 employees
3 yrs
To first real proof point
1M
CoCounsel users, early 2026
+11–14%
Stock move on that milestone

The lesson is not "buy AI companies." It's that the exposure was priced and acted on before it showed up in a quarter. Run this enterprise's numbers through the sliders above and ask the harder question: if you're the seller, would you rather be Casetext being bought ahead of the disruption — or the target nobody bought in time?

Sources: Thomson Reuters press materials and SEC filings, 2023; TechCrunch, June 2023; CIBC Research deal note, June 2023; MLQ.ai, February 2026.