A multiple without a starting number
Start with the number Salesforce leads with, because it cannot be checked from anything Salesforce published alongside it. Twenty-one times is a ratio, but Salesforce supplies neither the starting value nor a definition of the metric being multiplied. Separately, it reports more than 185,000 users. The release does not establish that this user count is the endpoint of the 21x calculation. Usage and users are not the same thing. Usage could mean tasks delegated, sessions, messages, or active days, and a small pilot cohort using a tool daily can multiply any of those quickly without the user base moving much at all.
That does not make the claim false. It makes it unusable for the decision it is meant to support. A buyer reading "21x in three months" is being asked to infer momentum, and momentum is exactly what a baseline would establish or puncture. Twenty-one times a hundred pilot users is a pilot. Twenty-one times fifty thousand is a rollout. Salesforce knows which one this is. The release does not say.
The user total has the same gap in the other direction. More than 185,000 users are "putting Coworker to work", in the release's phrase, but there is no split between paid seats, other Agentforce plans, Flex Credits use and trials. At the new price, the hypothetical arithmetic is straightforward: 185,000 users multiplied by $50 a month would equal $9.25 million in monthly billings. But that assumes every user takes the new per-user option, something Salesforce has not disclosed. The calculation is not revenue, an ARR estimate or evidence of paying adoption. What a customer count without a paying count measures is distribution, not demand.
What is actually available, and when
Set the growth claim aside and the release contains a concrete, staged product change. Four items are listed as available now, as of November 5: Google Workspace and Microsoft 365 connectivity, Conversation History, Skills, and the $50-per-user pricing option. The remaining announcements have later availability windows and should not be treated as shipped features. Those are the parts a buyer can test this week, and the connections matter more than the headline suggests. Coworker's pitch rests on context, and until now that context was mostly Salesforce's own records. Pulling in the email and files around an account is what turns a renewal brief from a CRM summary into something a seller might recognize.
The rest is a schedule, not a shipment. The Teams surface and the phone and desktop apps, arguably the parts most employees would actually touch, are listed as generally available at some point in November, with no day given. Scheduled tasks and the Daily Digest, the two features that would make Coworker proactive rather than reactive, are expected in December. Salesforce's footer separately says pricing is subject to change and advises buyers to base purchasing decisions on currently available features. That advice is particularly relevant here: several of the most prominent additions are still scheduled for later release.
One more distinction is worth keeping. The Adecco Group's planned expansion of Coworker across 40 countries is a deployment announcement, not a measured outcome. The release offers no time saved, no cost avoided, and no quality measure from Adecco or anyone else, and Adecco's quoted comment is about having candidate and client context available at the moment of judgment. Access is a plausible benefit. It is also the easiest benefit to claim, because it is true the moment the integration is switched on, whether or not anyone's work improves.
The practical test
Coworker is not hard to evaluate, which is why the missing numbers stand out. A team can price it exactly, at $50 per user per month as of November 5, 2026, start with the features that are live now, and measure the one thing Salesforce did not publish: whether delegated jobs are completed accurately enough, and require little enough correction, that reviewing the results costs less than doing the work directly. Until a vendor publishes a baseline, a usage definition and an outcome, its growth multiple is a claim about itself. Treat it as one.
Sources
- [1] Salesforce (primary), press release, "Salesforce's Agentforce Coworker Usage Soars 21x in Three Months as Enterprises Delegate Real Work to AI", November 5, 2026, Singapore edition (read first-hand in full at Gate 1 via salesforce.com/ap: 21x growth in usage in three months, with no baseline, definition or method given; more than 185,000 users stated as a separate figure; US$50 per-user, per-month option; Availability section — available now: Google Workspace and Microsoft 365 connectivity, Conversation History, Skills, per-user SKU; generally available in November 2026: Coworker in Microsoft Teams, iOS app, Mac and Windows desktop apps, Coworker MCP server, free trial for Flex Credits customers; expected in December 2026: Daily Digest, scheduled tasks, Coworker in the Field Service, Consumer Goods Cloud and Life Sciences Cloud mobile apps; The Adecco Group is the only customer named in the body, expanding Coworker across 40 countries, its CEO quoted on access to candidate and client context when human judgment matters most; the footer states pricing and packaging are subject to change and advises customers to base purchasing decisions on currently available products and services; no paying-user count and no productivity or savings figure appear anywhere in the release)Read source
- [2] Forkast, "Salesforce Expands Coworker's Reach, but Its 21× Growth Claim Has No Baseline", November 5, 2026 (independent check making the same missing-baseline point; cited for corroboration of the gap, not for product facts)Read source
- [3] CX Today, "Can Salesforce Win the AI Front Door Inside Teams?", November 5, 2026 (secondary coverage of the Teams move and the $50 option; product facts in this article are taken from Salesforce's release, not from this piece)Read source