On 3 September 2026, Salesforce scrapped its per-product pricing model. In its place: three bundled editions, Core, Advanced and Max, that fold Agentforce, Slack, Tableau Next and security features into flat per-seat tiers priced at $195, $395 and $550 per user per month.
Each tier also comes with a pool of what Salesforce calls Flex Credits, its consumption unit for AI and agent usage, set at 500,000, 1,000,000 and 2,750,000 respectively. That structure is the interesting part. Salesforce isn't charging per agent action, and it isn't pricing AI as a separate line item on top of the seat. It's bundling AI usage into the seat price and backstopping it with a shared credit pool.
What Salesforce actually changed
Under the old model, Agentforce, Slack and Tableau Next were sold and priced as distinct products, each with its own metering logic layered on top of a customer's core Salesforce seats. According to Salesforce's own announcement, and reporting from Salesforce Ben and CIO.com, that's gone. Core, Advanced and Max are now the only three editions, and moving up a tier gets you more of everything: more Slack functionality, more Tableau Next capability, and a much larger Flex Credit allowance for Agentforce usage, rather than requiring a separate purchase decision for each product.
The credit pool scaling is worth sitting with: 500,000 credits on Core, doubling to 1,000,000 on Advanced, then nearly tripling again to 2,750,000 on Max. That's not a linear step between tiers. It's a curve designed to make the top tier look disproportionately generous on AI usage specifically, which tells you where Salesforce expects the real usage growth, and the real willingness to pay, to come from.
Why this is a signal, not just a Salesforce story
Every SaaS founder currently shipping an AI feature is wrestling with the same pricing question: per-seat, usage-based, or some hybrid. Pure usage-based pricing captures cost accurately but makes customers' bills unpredictable and finance teams nervous. Pure per-seat pricing is predictable but leaves money on the table when a small number of power users drive most of the AI cost. Salesforce, at a scale where it can run this pricing experiment across its entire customer base, has picked a third option: flat per-seat tiers with a usage-credit pool bundled in.
That's a meaningful data point precisely because Salesforce isn't guessing. It has the billing data on how Agentforce is actually being used across its install base, and it chose to abstract that usage into credits rather than either metering every action individually or ignoring usage entirely. When a vendor with that much pricing telemetry moves away from granular metering, it's worth asking why. The likely answer is that customers dislike unpredictable AI bills more than they mind a slightly generous credit pool.
What to actually borrow from this if you're pricing AI features
If you're currently metering every AI action separately, every completion, every agent run, every token, consider whether a credit pool would serve your customers better without giving up your own cost visibility internally. The mechanism is straightforward: keep your internal cost tracking exactly as granular as it is today, but expose a simple credit balance to the customer instead of a line-by-line usage bill. You still know your margin per action; your customer just sees they've used 40% of this month's credits, which is a far easier number to reason about than a variable invoice.
Size your tiers the way Salesforce did: don't scale the credit pool linearly with price. Salesforce's Max tier costs 2.8 times what Core costs, but it carries 5.5 times the Flex Credits. If your top tier is roughly three times the price of your entry tier, consider giving it meaningfully more than three times the usage allowance. That asymmetry is what makes an upgrade feel like an obvious move for a customer whose usage is growing, rather than a marginal one.
And decide now, explicitly, what happens when a customer exceeds their credit pool. Whichever way you handle it, hard cap, auto-upgrade prompt, or per-credit overage billing, say so clearly on your pricing page before a customer hits the ceiling and finds out by surprise.
The bottom line
Salesforce replaced per-product pricing with three bundled tiers, $195, $395 and $550 per user per month, each carrying a Flex Credit pool that scales faster than the price does. If you're still deciding how to price your own AI features, treat this as evidence that a bundled seat price with a usage-credit pool is winning out over granular per-action metering, and size your own credit tiers with the same non-linear curve.