Chapter 2

Expert AI

The share-price fall that opened this report is, at its core, a bet that generative AI erodes the value of Wolters Kluwer's proprietary professional content. This chapter tests that bet against the record. The evidence is that AI is so far extending the franchise rather than commoditising it: adoption is measurable and early, the moat mechanism is content plus human expert validation that a general-purpose model does not replicate, and monetisation is real but still mostly a retention-and-upsell story, not yet a cleanly separable new revenue line.

From experimentation to a branded strategy

Wolters Kluwer's AI posture has moved in three deliberate steps. In 2023 it "stepped up experimentation with large language models" across dozens of use cases [1]. In 2024 it "rolled out many GenAI features, including enhanced search, summarization, Q&A, and virtual assistants" across all divisions [2]. In 2025 it consolidated the effort under one brand — "Expert AI" — and released it into flagship products, "such as UpToDate Expert AI and CCH Axcess Intelligence" [3]. By the end of 2025, management states, "nearly 70% of our digital revenues are from AI-powered solutions" [4].

Source: FY2023 Annual Report [5]; FY2024 Annual Report [6]; FY2025 Annual Report [7].

The direct question — how the company is positioned "vis-a-vis the pure AI players" — was put to incoming CEO Stacey Caywood on the FY2025 results, and her answer names the mechanism the whole thesis turns on: "What has always differentiated our business and stood the test of time, is our high quality, proprietary content and our deep domain expertise" [8].

What "Expert AI" is — and why a general model does not copy it

Expert AI is defined as "the generative and agentic AI embedded in Wolters Kluwer solutions… Grounded in our proprietary content and supported by 'expert-in-the-loop' oversight" [9]. Two inputs make it hard to replicate. The first is content the company controls: it describes "100% proprietary content databases" and a proprietary AI-enablement platform ("FAB") built in-house by a 6,500-strong technology team [10]. The architecture layers that proprietary content and customer data on top of primary sources, then routes the output through an "expert-in-the-loop" validation step before it reaches the user [11].

The second is the expert labour behind that validation. The company co-develops these solutions with thousands of domain experts: 7,600 clinicians and 100+ editors in Health; 800+ tax analysts; 500+ in-house legal experts working with 35,000 external legal authors; plus experts in compliance and ESG [12]. This is the crux of the durability case: a foundation model can summarise text, but it does not employ 7,600 clinicians to debate and validate what the guidance should say. The company's own framing is that this validation is what "performs much better than general-purpose AI models that lack content or domain knowledge" in the high-stakes decisions its products serve [13].

That claim shows up in the retention numbers. Renewal rates "for our largest subscription-based expert solutions… remained above 90%" through 2025 — the year Expert AI launched into the base [14]. The honest caveat, carried forward from the opening chapter, is that renewal is a lagging indicator: it held above 90% before AI-native competitors reached scale in these professions, so it evidences resilience to date, not immunity.

Adoption is measurable — and early

The most useful evidence is not the strategy but the take-up, and here the disclosure is unusually concrete. UpToDate is roughly 10% of group revenue and about 99% recurring, so its AI transition is a fair test of whether customers pay for the upgrade [15]. Since the October 2025 launch, roughly 80 enterprises had signed for UpToDate Expert AI — including 30 of the top 100 enterprise customers — and about one-third of renewing individual subscribers were upgrading to the premium Pro Plus bundle [16].

UpToDate Expert AI take-up since the October 2025 launch. Source: AI Investor Teach-In, 8 Dec 2025 [17]; FY2025 results call [18].

On the results call the Health leader put the enterprise number in sharper relief: "a third of our large health system customers, together representing some 1,600 hospitals, have signed up to take Expert AI" [19]. The same passage carries the sharpest counter-fact in the chapter: those signed systems "include health systems that are piloting tools from other LLMs or medical AI vendors" [20]. The competition is not a future threat; it is already inside the installed base. Management's read is that Expert AI clears "rigorous governance process and security reviews" that a general model does not, and that clinicians rate its answers highly — but the contest for the clinical-reference workflow is live now, not deferred.

In Tax & Accounting the equivalent evidence is product breadth: six Expert AI-powered modules launched across the CCH Axcess workflow, from document intake to conversational intelligence, on the industry's only fully cloud-native suite [21]. In Legal, the acquired Libra "Legal AI Workspace" launched across four European markets, combining the AI tool with proprietary content and workflow integration — the differentiation management draws against "standalone AI assistants" is "a single platform for research, analysis, and document creation, seamlessly integrated into existing workflows and customers' data" [22].

How it becomes revenue — real, but mostly upsell for now

The gap between "customers are adopting AI" and "AI is adding revenue" is where the durability case has to be tested hardest. Management lays out four monetisation avenues, and the honest reading of the table is that the weight still sits on the left-hand column — price increases that support retention — with the discretely-monetised and consumption-based models earlier in their life.

Source: AI Investor Teach-In, 8 Dec 2025 [23].

Two concrete pricing points give the upside texture. Libra, which moves the company into a new addressable market, sells at "about 2 times the value of the content offering that we would sell to a law firm" — a genuine price uplift, not a bundled giveaway [24]. And CCH Axcess Intelligence carries "consumption tiers", while the client-collaboration tool is priced against request lists sent — "a measure of output" [25]. But the same executive is explicit that for the 2023–2024 AI features, monetisation "was more about supporting our renewal rates with price increases" [26], and consumption-based pricing is "currently being tested" [27]. The company's summary line is that "AI innovation will drive our future organic growth" [28] — a forward claim the reported numbers have not yet had time to confirm as incremental rather than defensive.

The competitive reality: a shared playbook, not a unique one

The strongest structural caution is that content-plus-AI is not proprietary to Wolters Kluwer — it is the industry's answer. Thomson Reuters, which names Wolters Kluwer among its "primary global competitors," has "introduced agentic AI into our core offerings… backed by Checkpoint's comprehensive proprietary content and deep domain expertise from our tax editors" [29]. The same combination — trusted content, human experts, agentic workflow — is being deployed by the closest peers. That argues Expert AI is table stakes to defend the franchise, and the more speculative part of the thesis is whether it also expands it. (The RELX filing indexed here is a Netherlands financing entity rather than the operating LexisNexis business, so it cannot serve as a like-for-like operating comparison; the genuine peer read comes from Thomson Reuters and from Wolters Kluwer's own disclosures.)

There is also an option the company flags on the other side of the ledger: its Supervisory Board "discussed the risks and opportunities regarding potential content licensing deals in relation to AI," while continuing to "carefully monitor potential threats and business disruption" [30]. Licensing proprietary corpora to AI developers is a potential revenue avenue that does not depend on winning the end-user interface — a hedge worth watching, and a reminder the company itself treats disruption as a live risk rather than a settled question.

The measured read

On the evidence, the derating looks more like a re-rating of fear than a verdict the operating record supports — but the case is genuinely two-sided, and what would settle it is observable. The bull's best facts are the validation moat (7,600 clinicians and expert-in-the-loop, which a general model does not reproduce), measurable early adoption (30 of the top 100 health enterprises, one-third of renewing individuals upgrading), and real price uplift where the product opens a new market (Libra at roughly 2× content pricing). The bear's best facts sit in the same disclosures: rival AI vendors are already being piloted inside Wolters Kluwer's own accounts, the same content-plus-AI playbook is available to Thomson Reuters and others, and most AI monetisation so far is retention support rather than a cleanly incremental line, with consumption pricing still in test.

What would change the read, in either direction, is checkable in the filings to come: whether cloud and expert-solutions organic growth accelerates as Expert AI scales (the company's "drives future organic growth" claim made good), whether renewal rates hold above 90% as AI-native competitors mature, and whether the discretely-monetised and consumption-based avenues grow from a footnote into a disclosed contributor. Those line items — not the AI narrative — are where the durability question gets answered.