1. Home
  2. Case studies
  3. How Salesforce Needed Three Pricing…
Chapter 4 of The Subscription Playbook

How Salesforce Needed Three Pricing Models to Sell AI Agents

The company that wrote the blueprint for SaaS couldn't settle on one price for AI agents. So it offered several, and let customers pick their risk.

By Robert Coorey Updated 5 min read
The short answer

Salesforce launched Agentforce in October 2024 at $2 per conversation. In May 2025 it added Flex Credits at $0.10 per action, kept per-user licenses for employee AI, and later added a $2-per-resolution agent and fixed-price enterprise deals. By August 2026 Agentforce ARR passed $1.5 billion. When value is uncertain, offer two or three meters and watch which wins.

Key takeaways

  • No single price fits a new product whose value buyers can't yet judge.
  • Different buyers want different risks: predictability (per user), flexibility (per action) or alignment (per outcome).
  • Prepaid credit packs bring cash in early. By late fiscal 2026, more than half of Agentforce bookings were reportedly credit refills.
  • Too many meters scare buyers too. Fixed-price enterprise deals grew fast for exactly that reason.
  • Review a new service's pricing within 6 to 12 months. Salesforce changed course in seven.

When I wrote The Subscription Playbook, I called Salesforce "the blueprint for SaaS."

Before Salesforce, companies like SAP and Oracle sold CRM as standalone software they installed for you. Salesforce put it in the cloud for a subscription fee. Every software company since has copied some version of that model.

So here's what I find fascinating. When Salesforce launched AI agents, the company that wrote the blueprint couldn't make one price work.

It ended up with at least three.

How Did Salesforce First Price Agentforce?

Agentforce is Salesforce's family of AI agents. They answer customers and do work inside Salesforce.

When it became generally available on 29 October 2024, the price was simple: $2 per conversation.

Simple to understand. But it had two problems:

  1. It priced out smaller organizations and nonprofits.
  2. It felt expensive for quick interactions. A ten-second "what are your opening hours?" chat cost the same as a long, complicated one.

Sound familiar? It's the same flaw Cursor and Replit hit. A flat price per interaction doesn't match how much work each interaction takes.

What Did Salesforce Change?

It started adding meters. Fast.

Date Pricing change
29 Oct 2024 Agentforce launches at $2 per conversation
15 May 2025 Flex Credits: $0.10 per action (20 credits per standard action, $500 per 100,000 credits)
17 Jun 2025 $125 per-user add-ons for unmetered employee use
Oct 2025 Fixed-price Agentic Enterprise License Agreements (AELAs)
25 Jun 2026 Help Agent at $2 per resolution, with no charge for escalations
3 Sep 2026 New Core, Advanced and Max editions at $195, $395 and $550 per user per month, with bundled credits

Along the way, the credit rate cards were revised three times (October 2025, February 2026 and April 2026). SaaStr summed it up: Salesforce now had "3+ pricing models for Agentforce."

Salesforce also bought Fin (formerly Intercom), whose AI agent charges $0.99 per resolution. MarTech reported the deal in June 2026, and it closed on 10 September 2026. So the company now owns both a per-action meter and one of the best-known per-outcome prices.

The Numbers: Did Multiple Meters Help?

The flexibility appears to have helped adoption.

In its Q2 fiscal 2027 results on 26 August 2026, Salesforce reported:

  • Agentforce ARR above $1.5 billion, up more than 240% year on year (now including Slackbot and other AI offerings);
  • combined Agentforce and Data 360 ARR of nearly $3.9 billion, up more than 210%;
  • full-year revenue guidance raised to $46.1 billion to $46.4 billion.

Agentforce ARR reportedly grew from $800 million to $1.5 billion over three quarters. More than half of Agentforce bookings came from credit refills by late fiscal 2026, according to one pricing analyst.

There's a caution, though. SaaStr estimated in September 2026 that only about 3% of Salesforce customers ran Agentforce in production. Revenue is growing fast from a small slice of a very big customer base.

Which meter suits your AI agent?

Compare per-seat, per-conversation and per-outcome pricing on your own numbers in the AI Agent Pricing Calculator.

Open the free tool

Why Did Salesforce Need So Many Prices?

Because buyers of a brand-new product don't agree on what risk they want to carry.

Meter What the buyer gets Who it suits
Per user Predictability Finance teams who hate surprises
Per action (Flex Credits) Flexibility Teams experimenting with lots of small tasks
Per outcome (Help Agent) Alignment: pay only when it works Buyers who want proof first
Fixed-price AELA No meters at all Large enterprises that want one number

None of these is "right." Each shifts risk between Salesforce and the customer in a different way. Offering several lets each buyer choose. Our comparison of seat-based vs usage-based pricing goes deeper on the trade-offs.

But there's a limit. AELAs grew from 16 to more than 120 in a single quarter, according to the same analyst, and a Salesforce executive said European customers were "afraid of all these metrics." Choice helps. Too much choice sends buyers running for a fixed price.

How Does This Connect to the Book?

Two ideas from the book jump out.

Switching costs protect the experiment

In the chapter on building a moat, I wrote that CRMs like Salesforce secure customers through what I call the learning trap. Moving to another CRM means migrating data and retraining staff. That moat gave Salesforce room to experiment with agent pricing without losing its core customers.

Sales compensation gets harder

In the book I told the story of two salespeople, John and Anna. John made 25 small sales. Anna made five sales to businesses that would grow their usage enormously. Under usage pricing, Anna deserves more commission.

Now imagine running a sales team selling four different Agentforce meters. Paying people fairly gets complicated fast. That's one hidden reason companies drift back toward fixed-price deals: they're easier to sell and easier to pay commission on.

My take

I don't think three meters is the end state. It's what a market looks like before anyone knows what an AI agent is worth. My bet is that Salesforce will keep one simple meter for each type of buyer and quietly retire the rest. Watch which one the refills come from.

What You Can Steal From Salesforce

  1. Offer two ways to buy a new service. A fixed monthly retainer next to a pay-per-job rate. Customers self-select by risk appetite.
  2. Sell prepaid credit packs. You get cash upfront and the customer gets flexibility. The Credit Pack Designer helps you set pack sizes and margins.
  3. Keep a "no meters" option for big clients. Some buyers will pay more for one predictable number.
  4. Revisit new-service pricing after 6 to 12 months. Salesforce changed course within seven months of launch.
  5. Don't let options multiply forever. If nobody picks an option, drop it.

Not sure which model suits you? Try the Pricing Model Picker, or read our full guide to AI agent pricing. For the per-outcome side of the story, read how Intercom bet the company on 99 cents.

Frequently asked questions

What are Salesforce Flex Credits?

Flex Credits are prepaid units that Agentforce agents consume as they work. They're sold at $500 per 100,000 credits, and a standard action uses 20 credits, or $0.10. Customers buy packs upfront and refill them. Salesforce introduced them on 15 May 2025 so customers could pay for what an agent actually does rather than a flat fee per conversation.

How much does Agentforce cost per user?

As of October 2026, Salesforce lists several per-user options: an Agentforce User License at $5 per user per month plus Flex Credits, Agentforce add-ons for Sales, Service or Field Service at $125 per user per month for unmetered employee use, and new editions from 3 September 2026 at $195, $395 or $550 per user per month with bundled credits.

What is an AELA?

An Agentic Enterprise License Agreement is Salesforce's fixed-price, multi-year contract that gives unlimited use of named products. It exists for buyers who don't want to track meters at all. According to one pricing analyst, AELAs grew from 16 to more than 120 in a single quarter, and a Salesforce executive said European customers were "afraid of all these metrics."

Is Agentforce actually being used?

Revenue says yes: Salesforce reported Agentforce ARR above $1.5 billion in August 2026, though that figure includes Slackbot and other AI offerings. Usage depth is less clear. SaaStr estimated in September 2026 that only about 3% of Salesforce customers ran Agentforce in production. Treat that as a third-party estimate rather than a company figure.

Should a small business offer more than one pricing model?

Two options can work well, for example a fixed monthly retainer next to a pay-per-job rate. Customers pick the risk they're comfortable with. More than three options usually confuses people. Start with one simple main offer, add a second for buyers who want the other kind of risk, and drop whichever one nobody chooses.

Get both books free

The Subscription Playbook (ebook + audiobook) and Feed a Starving Crowd, instantly. In Australia? Claim the free hardcover.

Your details go to Subscription Playbook and our sponsor BPAG Accountants Pty Ltd, who may contact you about the book and, if you ticked the box, your review. Unsubscribe any time. Privacy policy.

Robert Coorey

Robert Coorey, MBA

Robert is the author of The Subscription Playbook and Feed a Starving Crowd and a Director at BPAG Accountants in Sydney. More about Robert.

Free toolsFree book