The pricing reset: how MCP is about to break the expert network per-call model

Guidepoint, Third Bridge, and AlphaSense have all connected their transcript archives to Anthropic's Claude through MCP over the past several weeks, with Guidepoint exposing more than 100,000 compliance-reviewed transcripts to buy-side analysts working inside the model. The connector itself is plumbing. The pricing consequence is not.

Our read: once a hedge fund analyst can interrogate a six-figure transcript library from inside a chat window, the unit of consumption that has anchored the expert network P&L for two decades, the billable consultation, stops being scarce. Per-call pricing at tier-1 ENs is structurally over within 18 months. The firms that reprice first to library-and-seat will capture buy-side budget that used to be split across three subscriptions. The firms that don't will watch ARPU compress.

What just changed

The technical change is narrow. MCP lets an authorized client (in this case Claude) call a structured tool, in this case a search-and-retrieve interface to a vetted transcript corpus, without the analyst leaving the model. The compliance layer that expert networks have spent twenty years building, the redaction of MNPI, the speaker vetting, the consent capture, sits on the EN side of the connector. The analyst sees only what compliance has cleared. Guidepoint, Third Bridge, and AlphaSense have each shipped a version of this, with Guidepoint's insights archive the largest exposed surface to date at more than 100,000 transcripts.

The commercial change is wider. The buy-side workflow that used to require a moderator, a calendar, and a 60-minute call now starts with a query against a corpus that already contains 18 months of comparable conversations. The call is no longer the first step. It's the escalation when the corpus runs out.

How the historical model is priced

A standard buy-side subscription at a tier-1 EN bundles a fixed annual call quota, typically in the 40 to 150 range depending on fund size, with overage billed at roughly USD 1,000 to 1,500 per consultation. Transcript access has historically been a secondary line: included as a teaser in the base subscription, gated behind the call quota, or sold as a separate product (the Tegus model before AlphaSense bought it).

The economics work because the call is the scarce unit. The expert's hour is the inventory. The moderator's compliance review is the cost of goods. Every additional call is incremental revenue against a roughly fixed compliance team.

MCP breaks the scarcity. A 100,000-transcript library answers most of the questions an analyst would have used a call to answer first-pass. The call survives as a product, but only for the residual: the question the corpus doesn't already contain, or the speaker the corpus doesn't already cover. That's a smaller volume of calls at a higher per-call value, not the same volume at the same price.

Three scenarios for what reprices

Base case (most likely). Tier-1 ENs (AlphaSense, Guidepoint, GLG, Third Bridge) move to library-and-seat pricing within 12 to 18 months. Per-seat ACVs rise (because the library is now the primary product) but the per-call line compresses sharply. Net effect on tier-1 ARPU is roughly flat to modestly positive, with margin expansion as compliance scales against queries rather than calls. The buy-side consolidates from three EN subscriptions to one or two, because the library overlap makes the third redundant.

Bull case for incumbents. ENs with the deepest libraries (AlphaSense post-Tegus, Guidepoint) capture disproportionate share of the consolidated buy-side budget. AlphaSense's Tegus acquisition looks prescient in retrospect: USD 930M for the library that becomes the primary product, not the secondary one. Smaller ENs (Dialectica, ProSapient, Coleman) lose seats to the consolidators and either get acquired or retreat to specialist verticals where library depth matters less than expert recruitment.

Bear case for incumbents. AI-native entrants (Ethos, Bridgetown Research, others not yet visible) win the seat war because they were never structured around a per-call P&L. They can price the library at a 40 to 60% discount to incumbents because they don't have a moderator bench or a per-call revenue line to defend. The incumbents reprice anyway, but a quarter to two quarters later than they should have, and the gap is permanent.

The MiFID II analogue

The closest historical comparison is sell-side research after MiFID II unbundling in 2018. Before unbundling, research was paid for through bundled commissions; the unit of consumption was implicit. After unbundling, asset managers had to write a check for research as a discrete line item, and the question stopped being "is the research good" and became "is it worth the line item." Sell-side research ARPU compressed roughly 20 to 30% over the following three years, and the firms that repriced first (Autonomous, Redburn, a handful of boutiques) captured share from the bulge-bracket desks that held the line on bundled economics.

MCP does to expert networks what MiFID II did to sell-side research. It makes the unit of consumption legible, and once it's legible, it gets repriced.

What to ask an expert next

• For the EN heads of product: what's the corpus-to-call ratio in your buy-side accounts today, and where do you expect it in 12 months?
• For the buy-side: how many of your last 20 expert calls would have been answered by a sufficiently deep transcript library search?
• For compliance: how does MNPI review scale when the query volume is 100x the call volume?
• For the AI-native entrants: what's the seat price that makes the library product defensible against a free-tier ChatGPT plus a Bloomberg terminal?

Watch the next two earnings cycles for AlphaSense and any disclosed pricing changes at GLG and Third Bridge. The first tier-1 to publicly move to library-and-seat sets the anchor everyone else negotiates against.