Salesforce Put Its CRM Inside Claude: The First Operating Proof That AI Needs a Software Partner

For most of 2026, Salesforce was the poster child for the fear that artificial intelligence would hollow out the software business. The story — the "SaaSpocalypse" — went roughly like this: if a customer can ask an AI agent to update records, draft emails, and manage a pipeline from inside a chatbot, why keep paying for the dashboard at all? By February the fear had helped wipe out roughly $2 trillion in software market value, and Salesforce, still down about 60% from its December 2024 peak, was off 22% for the year even as the S&P 500 climbed 12%.

Then one August day the company did something that made that thesis suddenly look backwards.

On August 26 Salesforce reported its fiscal second-quarter results and, on the same call, unveiled "Claudeforce," an expanded partnership with Anthropic, the company behind the Claude model. The product is deceptively simple: it puts Salesforce's data and workflows inside Claude's chat interface, shipping a plugin with 37 prebuilt sales skills so a seller can compose an email or update a record without ever opening a Salesforce screen. It is the first time the company has ever attached its "force" suffix to another company's product.

And it is where the quarter starts to prove a point about how AI actually reaches a business.

The SaaSpocalypse story assumed the model could replace the software layer. Claudeforce says the opposite: a frontier model can reason, but it has nothing trustworthy to reason over — no governed records, no business rules, no permissioning — unless it is wired into the system that owns them. Marc Benioff put it bluntly, telling CNBC that "frontier models depend on CRM. They don't replace it."

For the first time, that is more than a claim — it shows up in the numbers that determine near-term revenue.

Salesforce reported $33.5 billion in current remaining performance obligations — cRPO, roughly the contracted revenue it expects to recognize over the next twelve months — up 14% year over year, and called its net new annual order value the strongest in four years. cRPO is the metric to watch here because it is contracted and near-term, the closest a software business comes to delivered demand.

Within that, the AI line finally left the demo stage. Agentforce and Data 360, the agent and data products, reached roughly $3.9 billion of annual recurring revenue, up more than 210% year over year, with Agentforce itself above $1.5 billion, up about 240%. Salesforce delivered 3.2 billion "Agentic Work Units" in the quarter, up 97% sequentially, and raised its full-year revenue forecast to $46.1–46.4 billion. The stock jumped about 12–14% after hours and roughly 25% over the following week.

Now the discipline, because not everything in that headline was operating.

A large share of the earnings beat was a paper gain. Salesforce reported adjusted earnings per share of $5.90, a striking jump — but roughly $2.53 of that came from marking up its early stake in Anthropic, an investment now valued around $965 billion after Anthropic's May funding round. Underlying operating earnings were closer to $3.37 a share. The mark is a real strategic asset; it just is not a reflection of how the product business ran the quarter.

Second, the AI growth figures sit on a widened definition. Starting this quarter, the Agentforce and Data 360 ARR number was expanded to include Slackbot and the "headless" offerings. That makes the 210% and 240% growth rates directionally useful but not strictly comparable to a year ago.

Third, and most important for the near-term return: the absolute scale is still small, and the overall growth has not reaccelerated yet. Agentforce's roughly $1.5 billion of annualized revenue sits inside a company on track to do more than $46 billion this year. Total revenue grew 11%, and the full-year guidance raise to $46.1–46.4 billion is modest — about half of it traced to pending acquisitions of Contentful and Fin, not organic demand.

So the quarter proves one thing and not another.

It proves the bear case misread the architecture: agents do not bypass the enterprise system of record, they plug into it — and a model vendor with an estimated $65 billion revenue run rate found it worth partnering with the software owner to get there. That is why Salesforce's decision to wire its CRM into Claude, rather than hide from it, is the strongest operating sign yet that AI needs a software partner instead of a replacement.

What it does not prove is that the agent era is ready to reaccelerate Salesforce's growth in the way the rebound asks. At about $257, the stock trades for roughly 19 times trailing EBITDA — a premium multiple for an 11% grower. The market is paying for a conversion that has not reached revenue yet. The evidence is in bookings and ARR, which is exactly where this thesis lands or falls: whether Agentforce turns its $1.5 billion-plus of annualized contracts into recognized revenue growth over the next few quarters. Demand is not the issue. The issue is whether the metric converts — and whether that conversion was already priced into the stock when it rose a quarter in a week.