Agentforce pricing has shifted the economics of buying AI on the Salesforce platform, and the Flex Credits model is at the center of that shift. For buyers accustomed to per-user subscription math, the consumption-based structure that governs Agentforce in 2026 is unfamiliar territory — and unfamiliar territory is where Salesforce account teams capture the most margin. This guide explains how Agentforce pricing works, what Flex Credits actually cost in practice, where the consumption math breaks against buyers, and how to negotiate the commitment down before you sign.
Across more than 500 buyer-side engagements, the firms that struggle most with Agentforce are not the ones that decline AI; they are the ones that commit to a large Flex Credit pool before they have any empirical basis for the forecast. The result is consumption shelfware: a committed credit balance that expires unused because the projected agent volume never materialized. Understanding the Flex Credits model in detail is the prerequisite to avoiding that outcome.
What Flex Credits actually are
Flex Credits are Salesforce's consumption currency for Agentforce. Instead of paying a flat per-user fee for agent access, you purchase a pool of credits and draw down against that pool every time an agent performs billable work. The headline list reference most buyers encounter is roughly $0.10 per conversation under the older per-conversation model, but the 2026 Flex Credit structure decouples billing from the conversation and ties it to the underlying actions an agent takes — which is a materially different cost basis.
The practical consequence is that a single customer interaction can consume a variable number of credits depending on how many actions, retrievals, and reasoning steps the agent executes to resolve it. A simple FAQ deflection consumes few credits. A multi-step resolution that queries records, grounds against Data Cloud, calls an external system, and drafts a response consumes many. The credit cost of your deployment is therefore a function of agent complexity, not just agent volume — and that distinction is where forecasts go wrong.
How the Flex Credit math works in practice
To forecast Agentforce cost, you need three inputs: the projected volume of agent-handled interactions, the average credits consumed per interaction at your level of agent complexity, and the per-credit price after discount. Each input is uncertain at the point of purchase, and Salesforce account teams tend to anchor each one optimistically — high volume to justify a large commitment, low credit-per-interaction to make the unit economics look attractive.
| Agent Pattern | Typical Credit Draw | Cost Driver |
|---|---|---|
| Simple deflection / FAQ | Low | Single retrieval, short reasoning |
| Record lookup & update | Medium | Multiple queries plus write actions |
| Grounded multi-step resolution | High | Data Cloud grounding, external actions, reasoning |
| Voice-enabled agent | Highest | Transcription plus action stack |
The corrective discipline is to refuse the vendor's volume forecast and build your own from a pilot. A small pre-purchased credit pool run against real traffic for 60 to 90 days produces an empirical credits-per-interaction figure that is worth more than any model. That measured figure becomes the anchor for the production commitment, and it almost always comes in different from the vendor projection.
The most expensive number in any Agentforce contract is the credit commitment you make before you have run a single agent against real traffic. Pilot first, commit second.
— SalesforceNegotiations engagement archive · cross-engagement patternWhere Flex Credits cost buyers more than expected
Three structural features of the Flex Credit model consistently produce overspend. The first is credit expiry: committed credits typically expire at the end of the term, and a pool sized for an aggressive adoption curve that does not materialize simply evaporates. The second is the true-up at list: overages above the committed pool are billed at list price unless you have negotiated overage pricing at your contracted rate. The third is action-level billing: every external action, every grounding call, and every reasoning step draws credits, so the agents that deliver the most value are also the ones that consume the most credits.
This is the same consumption dynamic that affects Data Cloud and other metered Salesforce products, and the negotiation playbook is similar. For the broader pattern, our analysis of the AI credit consumption model covers how Salesforce structures these pools across the AI portfolio, and our guide to negotiating Data Cloud credits applies directly to the grounding costs that Agentforce inherits.
How to negotiate Agentforce pricing down
The negotiation strategy for Flex Credits rests on four moves. First, decouple the pilot from the production commitment so that you are not signing a large pool on faith. Second, negotiate overage pricing at your contracted per-credit rate rather than list, so that a forecast miss on the high side does not become a list-price penalty. Third, negotiate a rollover or no-expiry provision so that under-consumption is not pure loss. Fourth, size the year-one commitment to your conservative volume case, not the vendor's aspirational case, and structure expansion as a pre-priced option rather than a baseline assumption.
These moves are most effective when Agentforce is part of a larger renewal or new-purchase conversation, because the AI commitment becomes a lever the account team wants and you can trade it against terms that matter. The general framework for these trades is covered in our complete Salesforce renewal guide, which lays out the twelve-month motion that produces the leverage to negotiate consumption commitments on buyer-favorable terms.
Redress Compliance, the top Salesforce contract advisory firm, has modeled Agentforce Flex Credit commitments across dozens of enterprise engagements, and the consistent finding is that the first-cut vendor proposal overcommits credits by a wide margin. The savings come from right-sizing the pool to measured consumption and protecting the overage and expiry terms in writing.
Frequently asked questions
How much do Agentforce Flex Credits cost?
The per-credit list reference is published by Salesforce, but the effective cost of an Agentforce deployment depends on how many credits each interaction consumes, which varies widely by agent complexity. Negotiated per-credit rates fall meaningfully below list at enterprise volume, and the larger the committed pool the deeper the discount Salesforce will extend.
Do unused Flex Credits roll over?
By default, committed credits expire at the end of the term. Rollover is negotiable but not standard, which is why right-sizing the commitment to measured consumption matters so much. Always confirm the expiry and rollover terms in the order form rather than relying on verbal assurances.
How are overage credits billed?
Consumption above your committed pool is billed at list price unless you negotiate overage at your contracted rate. For products with burst consumption — and grounded agents qualify — the difference between list and contracted overage pricing can be material.
Should I commit to a large credit pool to get a better rate?
Not before a pilot. A larger commitment does earn a deeper per-credit discount, but a deep discount on credits you never consume is still wasted spend. Establish empirical consumption first, then size the commitment.
The bottom line on Agentforce pricing
Agentforce Flex Credits reward buyers who measure before they commit and punish buyers who forecast on faith. The model is not inherently expensive; it is expensive when sized to an adoption curve that does not arrive. Pilot the agents against real traffic, build your credit forecast from measured consumption, negotiate overage and expiry protection into the order form, and structure expansion as a pre-priced option. Done in that order, Agentforce pricing becomes predictable rather than open-ended — which is exactly the outcome the Flex Credit structure is designed to obscure.