Chapter 2

Retention and AI

Gartner's case rests on whether its subscription-research franchise keeps its clients when the questions those clients ask can increasingly be typed into a chatbot. The evidence points to a moat that is real but narrower than the headline suggests: enterprise client retention has held in a tight 83–89% band for six years, including through 2025's stall, while the number of client enterprises has shrunk and spending per retained client rolled over. The lock-in is durable; the expansion that used to sit on top of it is not guaranteed.

Sources: FY2025 Annual Report (Form 10-K), Item 1 Business [1] and Item 7 MD&A, Reportable Segments [2].

How the lock-in actually works

A Gartner subscription is not a library card. Clients sign contracts with a minimum term of twelve months that are generally non-cancelable and non-refundable, and at the end of 2025, 77% of contracts were multi-year [3]. Underneath the contract sits the mechanism that makes it stick: the published research is only the entry point. Licensed users get direct access — through inquiry — to a network of more than 2,400 analysts, and Gartner's experts logged more than 510,000 direct client interactions in 2025 [4].

Management is explicit that engagement and retention are linked: "Clients who engage frequently with our insights receive greater value and retain at higher rates" [5]. The switching cost is not a contractual penalty; it is the analyst relationship and the workflow built around it — the running advisory access a client loses mid-decision if it walks. That cost is best measured by revealed behaviour, and the behaviour is remarkably steady.

The retention signature

The clearest read on the moat is the retention series, split into its two components. Client retention answers "do they stay?" Wallet retention answers "do they spend more?" Gartner reports both for each of its two sales engines — Global Technology Sales (GTS), the larger book selling to technology buyers and vendors, and Global Business Sales (GBS), selling to HR, finance, supply chain and other functions.

Sources: FY2021 [6], FY2022 [7], FY2023 [8], FY2024 [9] and FY2025 [10] Annual Reports, Item 7 MD&A.

Two things stand out. First, client retention barely moves. GTS has renewed 83–86% of its enterprise clients every year for six years — through the COVID shock, the 2022–23 slowdown, and 2025's stall — and actually ticked up to 85% in 2025 from 84% [11]. An 85% renewal rate implies roughly 15% of enterprise clients leave in a given year; that number has been stable enough that it reads as the natural churn of a business embedded in customer workflow, not a franchise coming apart.

Second, what broke in 2025 was wallet retention, not client retention. GTS wallet retention fell to 96% from 102%, its lowest reading since 2020 and its first sub-100% level in five years; GBS fell to 99% from 106% [12]. Management attributes the decline to "lower levels of spending by existing clients," not to clients leaving [13]. The distinction matters for the durability question: the clients stayed and used the product more, but they bought fewer seats and add-ons than the year before. That is the profile of a demand air-pocket layered on a sticky base — consistent with the federal-spending shock and a cautious enterprise IT budget — rather than the profile of a product losing its reason to exist.

The engagement indicators cut both ways

The retention rates are reassuring; the volume metrics complicate the picture. Over five years the count of distinct client enterprises has fallen from more than 15,000 to just over 13,000, even as direct interactions climbed to a record 510,000 and the multi-year share of contracts rose to 77%.

Sources: FY2022 [14], FY2023 [15], FY2024 [16] and FY2025 [17] Annual Reports, Item 1 Business. Enterprise counts are disclosed as approximate ("over"/"close to").

Enterprise counts are disclosed only as round approximations, so the trend matters more than any single figure — and the trend is down by roughly two thousand logos in five years. The benign reading is deliberate concentration: management says its long-term strategy is to deepen its most valuable relationships, and rising interactions against a smaller base means the retained clients are engaging more, not less [18]. The less benign reading is that the low end is thinning — smaller technology vendors and marginal seats that a free web search or an internal AI tool can now cover — and that this is where the wallet-retention softness is showing up first. Both readings fit the same numbers; the corpus does not settle which dominates, and a widening logo count is the single indicator that would most cleanly confirm the base has stabilised.

The AI question

Gartner is candid about the competitive backdrop in a way worth quoting. In its own risk disclosure it notes that it competes with "free sources of information available to our clients through the internet," that "limited barriers to entry exist," and that it "anticipate[s] encountering more competition with increased adoption of AI services" [19]. It goes further in Item 1A: third parties "may be able to use AI to create technology that could reduce demand for our products," and although contractually prohibited, clients "may load our proprietary information into large language models, which could reduce the value of our offerings" [20]. This is the disintermediation fear stated by the company itself.

The offsetting facts are equally on the record. AI is, in management's words, "one of the most requested topics across all the roles we serve" — meaning the same technology that threatens the product is also driving demand for guidance on how to use it [21]. Gartner launched AskGartner, an AI interface to its own content, in August 2025 [22]. And asked directly on the Q1 2026 call whether Gartner would distribute its data through third-party LLM providers as some information peers have, CEO Gene Hall drew the line where the moat is: "Our published content is only a portion of what our analysts know — inquiries and analyst access unlock much more," so feeding proprietary data into an LLM "would not align well with our core value proposition, which is proactive, human-centered advisory" [23].

That claim is testable against the data already shown. If an LLM could substitute for Gartner, the first place it would show is client retention — clients would stop renewing. Instead client retention rose in 2025 while spending dipped, the pattern of a budget pause rather than a substitution event. The table below frames the debate on shared facts rather than sentiment.

Sources: FY2025 Annual Report, Item 1 Business [24] and Item 1A Risk Factors [25]; Q1 FY2026 earnings call [26].

What the evidence supports

On the moat playbook's own test — does the claimed advantage show up in numbers, and has it survived a shock — the answer is a qualified yes. Gartner's advantage is company-specific rather than merely a good industry: the combination of proprietary content, 2,400-plus analysts reachable through inquiry, a 100,000-executive peer network, and 40 years of brand is genuinely hard to reproduce, and it has held its client base through three separate stress tests since 2020. That reads as a narrow-to-wide moat: durable stickiness, evidenced by six years of stable client retention, sitting on top of an expansion engine that is cyclical and, at the margin, exposed to AI substitution at the low end.

The strongest fact against that read is the pairing the bears will press: a client base shrinking by roughly two thousand logos while, for the first time in five years, retained clients spent less — exactly what early structural erosion would look like before it reaches renewal rates. What would change the read in either direction is specific and checkable: GTS wallet retention returning above 100% and the enterprise count stabilising would confirm the cyclical thesis; a second year of sub-100% wallet retention accompanied by client retention slipping below its 83% floor would confirm the structural one. Those two line items, reported every quarter, are where the moat's durability will be settled — and they are the leading indicators the franchise and its selloff hang on.