Salesforce makes gutsy bet to win AI agent race

Salesforce (CRM) wants to sell AI agents to everyone, including companies that may never log into Salesforce.

This is a strategy the company has been pushing all year, and it could reshape how the software giant makes money for the next decade.

The plan rests on one idea:

Instead of charging businesses for every employee who uses its software, Salesforce wants to charge for the work its AI agents actually do.

Investors are not sold yet. Salesforce (CRM) stock trades near $165, down more than 8% over the past five days and off roughly 33% so far in 2026.

Why Salesforce is moving beyond charging per user

For 25 years, Salesforce sold software the same way: one license, one seat, one human.

That model breaks if AI agents start doing the work humans used to do. Fewer workers can mean fewer seats, which can mean slower revenue growth.

So Salesforce is building new ways to earn money that do not depend on adding more people.

More AI Stocks:

• Wall Street sees something in Micron that could reshape the AI trade
• Microsoft has bad news for a key AI partner
• Salesforce just beat estimates, one analyst still sees more pain

The spotlight falls on something Salesforce calls Headless 360, a way to plug its AI agents and data into software customers already use, even outside Salesforce’s own apps.

In plain terms, “headless” means giving access to Salesforce’s engine without needing the dashboard. A company can run Salesforce agents within its own tools and never touch the familiar Salesforce screen.

Alongside it is Flex Credits, a consumption model in which the usage bill grows based on the volume of tasks agents complete, not the number of employees logged in.

Gary Hershorn &sol Getty Images

The usage numbers Salesforce keeps pointing to

To prove demand is real, Salesforce leans on a metric it created called Agentic Work Units, or AWUs. One AWU is a single task that an AI agent completes.

The company has delivered 3.8 billion AWUs across Agentforce and Slack, up 111% quarter over quarter, according to its first-quarter fiscal 2027 results.

Usage inside Slack jumped even faster. Slack AI grew 347% quarter over quarter, driven by Slackbot, Salesforce reported on its Agentforce page.

The company also processed 28.6 trillion tokens in the quarter, up 152%. A token is the basic unit of data an AI model uses to process, understand, and generate information.

So, when Salesforce says it processed 28.6 trillion tokens, that’s a raw measure of total data volume the models churned through.

What the market punished, and why

Strong usage did not prevent a slide.

The stock fell about 8.89% over the past week, according to Yahoo Finance data.

The disconnect comes down to a gap between activity and durable revenue.

Related: Salesforce CEO delivers blunt message on AI agents

Agentforce annual recurring revenue has crossed $1.2 billion.

That is real money, but small inside a revenue base above $45 billion, which makes it hard to move the overall growth rate for now.

Salesforce also guided second-quarter revenue to a range that landed below Wall Street’s consensus, Reuters reported, giving sellers a reason to short the stock.

Then there is the fear of competitors:

Investors worry that AI platforms from OpenAI and Anthropic could chip away at demand for Salesforce’s traditional CRM seats before the new model scales.

How the pricing shift connects to layoffs and a billing deal

Salesforce acquired m3ter, a usage-based billing platform, on June 8, according to an article by EBC Financial Group.

You cannot charge by consumption at scale without the necessary infrastructure to meter and invoice it, and m3ter supplies that.

In early June, Salesforce cut 86 roles across MuleSoft, Marketing Cloud, and non-core Agentforce functions, according to a California WARN filing first reported by Business Insider. That followed a larger February round of close to 1,000 jobs.

Put together, the billing acquisition and the layoffs point to a company rebuilding its cost base and its product stack around charging for AI by the task done.

What this means for investors watching CRM

There's good logic behind Salesforce’s strategy; however, it is not yet certain that it will pay off.

A few signals will tell you if the strategy is working:

What to track on Salesforce’s AI bet

• Consumption revenue scale: Watch whether Flex Credits and Headless 360 grow large enough to lift the total growth rate.
• Existing-customer expansion: More than 50% of Agentforce and Data 360 bookings came from current customers in Q1, according to Salesforce. A rising share signals stickiness.
• Guidance direction: Salesforce raised its full-year fiscal 2027 revenue outlook to between $45.9 billion and $46.2 billion. Another raise would help close the credibility gap.

The risks are also clear:

Salesforce earns less when customers buy fewer seats, and the risk is that seat losses will outpace the revenue from the new pay-per-task model.

If that happens, growth could stall before the new pricing gains traction.

Salesforce still has scale, entrenched customers, and a valuation below many software peers, which gives patient investors a cushion.

For now, Benioff has placed his bet. The market is waiting to see how it plays out.

Related: Goldman Sachs has blunt message for AI stock investors