Expert-Network Economics in the AI Era: What Actually Drives Call Quality in 2026

The expert-network category, dominated in the United States by the well-known set of providers that includes GLG, AlphaSights, and Third Bridge among others, was widely predicted through 2023 and 2024 to be under existential pressure from language models. The pitch of the challenger tools was that a summary of publicly available information about a topic, assembled quickly by an AI system, could substitute for the paid expert call that had been the industry's product for two decades. Two years on, the substitution has happened in the low end of the market, but the picture in the middle and upper end is different and worth describing precisely, because it changes what a serious research or investment function should be buying.

The first observation is that the specific use case AI has substituted well for is the background-orientation call, the one where a junior analyst wanted a forty-five-minute overview of an industry from someone who had worked in it, without a specific question to test. That call now happens against an AI system, faster and cheaper, and the volume the expert networks used to see for those calls has meaningfully declined. The networks have responded to this by de-emphasizing that segment of the business rather than pretending it is intact.

The second observation is that the specific use case AI has not substituted well for is the expert call where the buyer has a specific, non-obvious question that requires a person with direct, recent, first-hand operational experience to answer credibly. Investment analysts trying to verify a specific supply-chain fact, private-equity diligence teams testing a specific management-team claim, corporate strategy functions probing a specific competitive dynamic, and litigation teams needing subject-matter testimony are the specific segments where the expert-network product has held up and, in some cases, become more valuable rather than less. The reason is that AI-generated summaries can produce a plausible answer to almost any question, but they cannot, in the specific settings where the buyer will act on the answer with real money at stake, be trusted with the specific facts that a real operator would know.

The third observation is about pricing, and it is where the interesting shift is happening. The unit price of the average expert call has gone up in the segments where the product has held, because the value the buyer is extracting per call has gone up. The buyers are asking sharper questions, doing more preparation before the call, and using the call to test hypotheses that AI-assisted preparation has already helped them form. The networks that have adjusted well are the ones whose relationship managers understand that the buyer is now paying for the specific expert, not for the abstract service, and that the matching quality is now the primary driver of buyer retention.

The fourth observation is about the expert side of the marketplace. The specific experts who are in high demand in 2026 are the ones with recent, direct, verifiable operational experience in specific narrow domains. The generalist executive with broad industry experience is less differentiated than they were, because the AI-assisted summary can cover the general ground competently. The specialist who ran a specific function at a specific company in the last three years, and can speak to specific operational realities that are not in any public source, is the specific asset the networks are now competing to represent.

The fifth observation is about compliance and provenance, which has become a much larger part of the business than it used to be. The specific corporate and financial buyers who use expert networks now demand documented compliance workflows: which employer restrictions apply to a given expert, what topics are off-limits, what disclosures the expert has attested to, and what recording and retention protocols are in place. Networks that operated with lighter compliance postures in the past are now spending materially more on those functions, because the buyer will not engage without it and, in the specific case of enforcement actions against expert networks in earlier cycles, the buyers have institutional memory.

For a head of research inside an investment or corporate strategy function in 2026, the working guidance is that the expert-network line item is not a candidate for wholesale replacement by AI tools, but the composition of what it buys has changed. The specific questions being posed to experts should be sharper, the specific experts being requested should be more narrowly matched, the preparation before the call should be more thorough, and the throughput of low-quality background calls should have already declined. If the internal usage pattern still looks like the 2022 pattern, the function is likely both overpaying for the calls that AI could handle and underpaying for the calls that only a specific human expert can.

There is a broader point. The categories of professional-services work where AI has substituted at the low end and lifted quality at the high end, of which expert networks are one example, are showing a consistent pattern: the middle of the historical product mix hollows out, the top of the market becomes more valuable and more expensive, and the bottom of the market is unbundled. The organizations that adjust their buying to that shape are getting more research value per dollar. The organizations that continue to treat the category as a commodity are systematically getting the wrong end of the pricing curve.