Full narration of the briefing. Click a section heading to jump the player to that moment.
Salesforce now sells its AI in credits, and credits are where enterprise budgets go to get surprised. Agentforce bills in Flex Credits and conversations. Data Cloud burns its own credit pool. And the new agentic enterprise agreements pool all of it into one committed number with an uplift attached.
None of this is a reason to avoid the products. It is a reason to negotiate them like the consumption contracts they are. Here are the five things that decide whether the credit economy works for you or against you.
Tip one. Never let the three currencies blur into one number. Flex Credits run about 500 dollars per hundred thousand, and a standard agent action burns 20 of them. Agentforce conversations list at around two dollars each, falling to fifty cents to a dollar twenty at enterprise commitment, with bundles starting near one hundred eighty thousand dollars for a hundred thousand conversations a year.
And Data Cloud credits list around 32 cents, dropping to ten to eighteen cents in commitments above half a million dollars. Three currencies, three discount curves, three places to negotiate. A proposal that shows you one blended number is hiding two of them.
Tip two. Size the commitment from your own data. Across the agentic agreements we have reviewed, vendor forecasts ran thirty to fifty percent above actual first-year usage, and customers burned only forty-five to seventy percent of their committed pools. Credits mostly reset annually with little or no rollover, so every unburned credit is margin you donated.
The structure that works: a conservative base commitment sized from your own telemetry, plus a pre-agreed expansion rate for growth. Buyers who split the deal that way paid eighteen to thirty percent less over the term than buyers who committed to the forecast.
Tip three. Model the burn before you commit. On the Data Cloud side, one complex identity resolution job consumes fifty to two hundred credits, a large segmentation over a thousand, and unifying a million rows runs into the tens of thousands at the base tier. On the Agentforce side, one user request can trigger several actions, and credits are fungible across agents, Data Cloud operations, and bring-your-own-model prompts, which makes them flexible and hard to forecast in the same breath.
Run the burn model on your top five use cases, then negotiate rollover or true-forward rights so the credits you buy survive the year you bought them in.
Tip four. Cap everything that compounds. Agentic enterprise agreements typically carry annual uplifts of seven to twelve percent, compounding, and the generous first-term rate drifts back toward list at renewal. Four protections belong in the order form: uplift capped at CPI or lower, overage rates capped at your committed rate, the renewal rate locked at signature, and volume-tier pricing that survives a true-down.
A three-year commitment with real minimums can earn twenty to forty percent off, but only the protections decide whether you keep it in year four.
Tip five. Keep the AI on its own paper. The flat-fee agentic agreements bundle Agentforce, Data Cloud, and more into one commitment, which is efficient right up until you want to change course. Negotiate the AI as a severable component: its own exit rights, its own proof-of-value milestones tied to measurable outcomes, and its own renewal terms, so a disappointing agent program never holds your core CRM renewal hostage.
Unlimited AI, like every unlimited before it, is still a number. Your job is to know the number before Salesforce does.
One last point. At Redress Compliance we negotiate Salesforce AI and Data Cloud agreements on a pure contingency basis. Our fee is 25 percent of what we save you. If we save you nothing, you pay nothing.
Before you commit to credits, let us run the burn model on your numbers. com.
Redress Compliance works on contingency: our fee is 25 percent of what we save you. Nothing saved, nothing paid. Independent, buyer side only, never vendor funded.
Talk to a Salesforce negotiator