Loading today’s stories
Invest1 distinct publisher3 min readPublished
Salesforce, FactSet, S&P Global and Moody's have all wired their data into Claude and ChatGPT, and only Moody's will name a charging method, which leaves per-seat contracts metering a unit that agents no longer respect.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
A per-seat licence meters consumption only because a human clicking through a browser consumes at human speed, and the dashboard is what keeps the human in the loop, so when Salesforce lets a rep pull the same customer record by typing at Claude [1] the seat stops standing in for anything: one hedge fund analyst pointing agents at a data product can multiply the queries they would otherwise run [13], and a fixed seat price divided by a rising query count is, by construction, a falling price per query [4]. That is the monetisation hole, or rather the more interesting version of it, which is that nobody has to fill it this quarter and everybody has to fill it before renewal. Three of the four connector owners in Semafor's account have no rate to quote: Salesforce released no pricing for the Claude integration [2], FactSet's Patrick Starling calls commercialisation "evolving" [4], S&P Global's Peter Licursi calls pricing "an evolution that we are actively discussing with our customers at all times" [5], and only Moody's names a method, billing against what it expects a customer's usage will be [6][3].
The adoption figures are direction, not rate. S&P Global says more than 500 customers have connected to its MCP [3]; no total customer count is given, so 500 does not convert into penetration [2], and the protocol itself is roughly two years old, introduced by Anthropic in 2024 and mentioned on more earnings calls every quarter since, per AlphaSense [8][1].
The cost side is where a subscription gets squeezed. The customer, not the vendor, carries the token bill that the connector generates [12], so the vendor's licence line now shares a budget with a variable cost it neither sets nor can discount [5], and a renewal held against a rising token invoice is a worse conversation than one held against last year's seat count. The confidence these firms project that the chatbot route will not cost them their legacy interfaces [17] is a claim about screens, not about billing.
Semafor's columnist reads it the other way: the SaaSpocalypse is overstated because proprietary data rather than a website is what makes these firms valuable, and you can vibe-code a scheduler but not S&P Capital IQ Pro [14]. That is right about value and quiet about price. Cristina Pieretti of Moody's frames the connector as self-disruption, preferable to being disrupted by someone else [7], which is the honest version of a vendor conceding the interface to keep the data.
This is probably wrong, but the way I would bet: the interface concession is cheap, the billing concession is expensive, and the sell-off earlier this year [9] was pricing the second one while the trade press covered the first. The counter-cases are real. Upwork gave its MCP server away free in August because a marketplace only earns on a completed transaction, making chatbots an acquisition channel rather than a product [15], which works for take-rate businesses and not for subscription ones. And if a vendor lands per-query pricing that beats its old seat revenue on the same account, with retention intact, my thesis is dead. There is no industry playbook yet [11]. The first published rate card will be the number to read.
Ranked by verification strength, evidence, and original report placement.
Salesforce announced "Claudeforce," a partnership with Anthropic that lets salespeople look up customer records and check on deals by typing requests to Claude instead of clicking through a Salesforce dashboard.
Salesforce did not release pricing details for its new Claude integration.
S&P Global said more than 500 customers have connected to the company's MCP as of last quarter.
Patrick Starling, who helps lead AI strategy at FactSet, said adoption of the company's connector has taken off "like a hockey stick" since launching last year, and that commercialisation is still "evolving" for FactSet.
S&P Global's Peter Licursi, who spearheads AI strategy at the company, described pricing as "an evolution that we are actively discussing with our customers at all times."
Moody's head of digital content and innovation Cristina Pieretti said it is still "pretty early", and Moody's currently charges based on what it expects a customer's approximate usage will be.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 28, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Claudeforce: Salesforce hands Anthropic the seller's screen and keeps the write path2 distinct publishers
invest
Google's bank agent can read FactSet. The licence, not the model, decides how much.1 distinct publisher
product
Salesforce's answer to the SaaSpocalypse: 11% growth, and a $2.6B mark on Anthropic1 distinct publisher
invest
Scalable Capital puts ChatGPT, Claude and Grok inside the European order ticket2 distinct publishers
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
On the record, but self-reported
Four executives are named and quoted — FactSet, S&P Global, Moody's, Upwork — which is more sourcing than most agent-integration stories carry. The catch is that every quantity belongs to the company stating it: S&P Global counts its own 500 connections, FactSet supplies its own "hockey stick," and Moody's describes its own billing. The one outside datapoint, AlphaSense's rising earnings-call mentions, measures how often executives say "MCP," not how much anyone deploys or pays.
Shipped and used, size unknown
These are live integrations, not roadmap slides: Claudeforce was announced this week, Upwork's server has been public since August, and S&P Global and FactSet both report growing connector traffic. What no one can say from this reporting is how big any of it is relative to the legacy interfaces it sits beside — there are no seat counts, no share-of-queries, no connector revenue. Real motion, unmeasurable magnitude.
Vendor confidence runs ahead of the meter
Semafor's own framing is sober — its verdict is that monetisation makes or breaks the pivot. The overstatement sits with the companies: they project confidence that chatbot access will not cannibalise their legacy interfaces while three of the four cannot name what it costs, and "disrupting yourself" is a reassuring phrase for a revenue model that has not been decided. The gap is modest rather than gaping because nobody in the story claims the money question is solved.
Announcements aimed at two audiences
Read these launches as investor communication as much as product news. Semafor ties the connector rush to a sell-off that punished software business models, which gives every vendor a reason to be visible inside Claude and vague about what it earns there. Upwork is candid that free access is a customer-acquisition funnel, Moody's frames pre-emptive self-disruption as strategy, and AlphaSense — the supplier of the growth statistic — sells the tracking of exactly this kind of boardroom vocabulary. Nothing here is concealed; it is all pointed.
Single account, unresolved core
One outlet, no corroboration, and the fact that matters most — what any of this costs — is precisely the fact left blank. The directional finding holds up well: connectors are live and pricing is unsettled, and the executives said so themselves. Anything sharper, such as how much revenue per query these firms are giving up, cannot be reached from what has been published.