Claudeforce Raises New Questions About Salesforce AI Costs

Salesforce and Anthropic’s new “Claudeforce” partnership represents a significant expansion of the relationship between the two companies. Claude is moving deeper into Salesforce, while Salesforce data and workflows are becoming directly accessible from inside Claude.

For IT procurement teams, there’s a lot to unpack. Pricing and packaging haven’t been published yet, so there’s still a lot we don’t know. But the details released so far point to a more complicated cost stack, greater reliance on consumption-based pricing and the potential for overlapping AI spend.

Procurement teams should start preparing now for the Claudeforce sales conversations that are likely headed their way.

What is Claudeforce?

Announced August 26, Claudeforce expands the Salesforce-Anthropic partnership in three areas.

1. Salesforce in Claude is the most notable development. A new plugin gives Claude access to live Salesforce data, workflows and governance, with 37 prebuilt sales skills covering activities such as meeting preparation, deal health and pipeline review. Sellers can query Salesforce data and perform certain actions without leaving Claude.
2. Claude in Salesforce embeds Claude more deeply across the Salesforce portfolio. Claude is being used as the default reasoning model in Agentforce Vibes and Agentforce Coworker and can be selected through Agent Builder and the Atlas engine.
3. Claude in Slack extends the relationship further, with Claude powering experiences including Slackbot, Claude Tag and Slack Code.

There’s also an important architecture consideration for regulated enterprises. Claude can run through Amazon Bedrock within Salesforce’s Trust Boundary, designed to keep model traffic inside Salesforce’s virtual private cloud.

These capabilities are still early. The Salesforce-in-Claude plugin is pilot-only, open beta is expected in September of this year, and additional skill sets are expected later in 2026.  

How will enterprises pay for Claudeforce?

Enterprise software pricing is steadily moving beyond the traditional per-seat model as AI agents become more active users of enterprise applications. Instead of paying primarily for named users, customers increasingly have to account for API calls, actions, inference and other forms of consumption, sometimes across multiple platforms.

Claudeforce illustrates how quickly that can get complicated. Current indications point to separate Salesforce and Anthropic charges, potentially layering Salesforce API or headless consumption and Claude inference on top of existing Agentforce credits and commitments. It’s very possible enterprises could be managing two vendors, two invoices and multiple consumption meters supporting increasingly overlapping capabilities.  

Full pricing and packaging have not yet been published, which makes it difficult to calculate the actual cost of running Claudeforce at scale. When those details emerge, customers will need to look beyond individual SKUs and rate cards and understand the combined cost of running a workflow across both the Salesforce and Anthropic ecosystems.

Watch for overlapping consumption

This will be one of the biggest areas for procurement teams to scrutinize.

An enterprise may already have Agentforce Flex Credits or other Salesforce AI commitments when it begins using Claude to interact with Salesforce data and execute Salesforce workflows. Customers will need clarity on whether Claude-powered activity draws against those existing entitlements or generates additional Salesforce and Anthropic consumption.

There are several questions worth pressing both vendors on:

• What exactly is being metered?
• Does a Claude-powered action consume existing Agentforce credits?
• Does the same interaction also generate Salesforce API consumption and a separate Claude inference charge?
• Are additional Salesforce products or editions required?

Until Salesforce and Anthropic publish the commercial details, we won’t know exactly how these meters interact. But procurement teams should understand the potential for overlap before Claudeforce moves from experimentation into meaningful production usage.

Consumption-based pricing also requires a different approach to forecasting. Seat counts are relatively straightforward to budget. Agent activity can grow quickly as enterprises expand use cases, users and autonomous actions, making strong usage visibility and controls increasingly important.

Explore Claudeforce pilot-to-production pricing before you commit

Claudeforce is still a pilot-stage offering, which makes any early pricing particularly important to scrutinize.

New products are often introduced with commercial incentives designed to encourage adoption. Those initial terms may bear little resemblance to what customers ultimately pay when usage scales or when the product reaches GA. Procurement teams evaluating a pilot should therefore understand the commercial path into production before making a larger commitment.

Ask what changes when the pilot ends. What will be metered in production? Are rates protected? Are there consumption caps and alerts? What are the overage rates? Will additional commitments be required? And how could Claudeforce adoption factor into the next Salesforce renewal?

Customers should address pilot-to-production pricing protections before usage scales rather than trying to negotiate them after Claudeforce is embedded in the environment.  

Permissions deserve attention, too

The Salesforce-in-Claude plugin operates through individual user permissions, while some Slack-based interactions may use dedicated integration credentials. Existing access controls therefore deserve a closer look before Claudeforce is widely deployed.

Over-provisioned access becomes more consequential when an AI agent can execute actions quickly and at the scale of enterprise usage. Enterprises should audit permissions before a pilot, start with read-only access where practical and expand write permissions deliberately.  

There’s also a longer-term sourcing consideration around model optionality. Claude is becoming deeply embedded across Salesforce and Slack while Salesforce itself becomes accessible from within Claude. Enterprises should understand what it would take, technically and commercially, to introduce or switch to another model later rather than assuming that flexibility will remain intact.  

What should IT procurement do now?

With pricing still unpublished, the immediate job is preparation.

Start by understanding your existing Salesforce AI entitlements and consumption commitments, as well as any current Anthropic relationship. Identify areas where capabilities or consumption could overlap. Review permissions. Determine how you’ll measure the total cost of a Claudeforce workflow across both vendors. And decide what pricing protections and flexibility you’ll want if a pilot eventually expands into enterprise production.

Salesforce and Anthropic are creating a deeply integrated AI ecosystem, and the cost implications could be significant. We expect to learn much more about pricing and packaging in the coming weeks. In the meantime, getting your current environment and commitments in order will put you in a stronger position when the Claudeforce sales conversations begin.

NPI can help you assess your current Salesforce and Anthropic position, identify potential overlap, and prepare for the pricing, packaging and negotiation decisions ahead. Contact us to learn more.