Loading today’s stories
Invest1 distinct publisher2 min readPublished
Fortune says the market wrongly buried Salesforce, Booking Holdings and IBM. The evidence offered is Agentforce's $1.5 billion of recurring revenue, spread across more than 30,000 contracts.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
Divide the headline number by the deal count and the win looks more ordinary. Agentforce's $1.5 billion of annual recurring revenue across more than 30,000 closed contracts averages under $50,000 per customer per year [1]. Fifteen-fold growth in eighteen months is a genuinely fast product ramp [3], and it is also a book of business whose typical contract is the size of a seat expansion rather than a re-platforming. Set against the software value the sector has shed, the recurring revenue is under a tenth of a percent [2]. Revenue and market value are not the same unit, but the scale gap tells you how much of the selloff this particular exhibit can carry.
There is also a mismatch between the rebuttal and the thing being rebutted. By Fortune's own account, the bear case has already left obsolescence behind and now runs on acquisition cost, pricing power and margin. What is offered against it is a revenue tally and a throughput statistic: more than 216 trillion customer records processed this year [7]. Records processed is volume. It says nothing about what Salesforce conceded on price to close those agent contracts, or what the gross margin on them looks like.
The sturdiest part of the argument is custody. Nobody, the piece argues, is going to extract decades of deal history and support tickets and re-post them inside a model vendor's store, and nobody would want a language model as the system of record for proprietary customer data [11]. That is persuasive about floors. It explains why the existing revenue does not walk out, not why the next dollar arrives at a better margin than the last one.
The drawdown figures deserve a second look too. Down roughly 20% this year and 40% from the high only reconcile if the peak sits about a third above where the shares started the year [4]. Whichever of the two numbers gets quoted decides how thoroughly the market looks to have written the company off, and the larger one is measured from a level the shares held briefly rather than a settled base.
Then there is the roster. The text supplied names Booking Holdings and IBM as fellow beneficiaries but attaches figures only to Salesforce [12]. A three-company thesis resting on one company's product line is a Salesforce call in sector clothing, and it is a call about data custody, priced by whoever ends up believing that custody converts into pricing power rather than merely into retention.
Ranked by verification strength, evidence, and original report placement.
Salesforce stock is down roughly 20% this year and 40% from its high.
Agentforce, Salesforce's AI agent platform, has gone from $100 million to $1.5 billion in annual recurring revenue within 18 months of launch.
Well over 30,000 Agentforce deals have already been closed.
Over the last year, approximately $2 trillion in software value has been destroyed on fears that AI will render many software businesses obsolete, in what has become known as the "SaaSpocalypse".
The same Fortune piece states there is no question that many high-flying technology winners will be under increasing competitive threat from autonomous AI agents, and that the list of companies that look vulnerable is a long one.
Analysts at Wells Fargo declared that "lower cost of intelligence increases value of incumbent data".
Follow any of these and your For You feed starts watching them — no settings page required.
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.
Single opinion column relaying vendor figures
All material comes from one publisher and one opinion piece. The operating metrics are company-reported and unverified in the supplied text, the central rebuttal claims (margin compression is wrong, data cannot be moved, security surpasses LLMs) are asserted without data, and two of the three named beneficiaries carry no figures at all. The quantitative core - ARR, deal count, record volumes, drawdown percentages - is specific and internally consistent, which keeps the score above floor.
Real but shallow enterprise footprint
Disclosed adoption is genuine and broad by count - more than 30,000 closed deals, $1.5bn ARR, Claude integrated as the premier agent, sharply rising record ingestion, and a million Slack users connecting outside agents within a month. Depth is the constraint: under $50,000 of ARR per deal on average looks like departmental pilots rather than agents displacing core workflows, and no retention, expansion or production-outcome data is disclosed.
Thesis outruns the proof offered
The claim is that a roughly $2tn repricing of software is wrong and that named incumbents will gain pricing power. The proof offered is one product line at $1.5bn of ARR - about 0.075% of the value said to be destroyed - averaging under $50,000 per contract, plus assertions about data moats and security. Booking Holdings and IBM are asserted as winners with no figures. The gap is clearly positive, though not maximal: the underlying adoption numbers are real and growing fast, and the piece does concede that many incumbents face genuine agent-driven threat.
Promotional framing, no disclosed positions
The supplied text is an opinion column advocating specific listed equities, quoting a sell-side analyst approvingly, praising the CEO and a $25bn buyback, and relying entirely on figures the advocated company disclosed. No author or publisher position disclosure appears in the supplied material. That combination is a substantial incentive load on the framing, even though the individual numbers cited are checkable against company reporting.
Moderate: one publisher, checkable numbers
Confidence is limited by single-publisher, single-source coverage and by the body being truncated before the Booking Holdings and IBM sections, so their absence of evidence is partly an artifact of the supplied text. It is supported by the specificity of the figures quoted and by arithmetic that can be performed entirely within the source, which makes the derived per-deal and ratio findings robust regardless of the thesis.
invest
Claudeforce: Salesforce hands Anthropic the seller's screen and keeps the write path2 distinct publishers
invest
Salesforce's $3.9bn agent line grew 210%, inside a company growing 11%1 distinct publisher
product
Slack Code makes coding agents taggable teammates. Your merge-approval policy is now overdue4 distinct publishers
invest
Salesforce's 14% cRPO is the best agent-demand read yet. Organic growth is 6.5%2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 26, 2026