After months of pushback and budget blowouts, Salesforce is finally rewriting the rules of engagement for Agentforce. Salesforce has introduced a significant overhaul to its Agentforce pricing structure today, moving away from the controversial $2 per conversation model toward a more granular action-based approach with its new “Flex Credits” system. The move comes after months of customer feedback highlighting concerns about cost predictability and scalability with the original pricing model.
What Was Announced
Today, Salesforce unveiled three major pricing innovations for Agentforce:
- Flex Credits: A new consumption-based model charging $0.10 per action (20 Flex Credits) rather than $2 per conversation. Credits are available in packs of 100,000 for $500, managed through Salesforce’s Digital Wallet, which provides usage analytics and forecasting tools.
- Flex Agreement: A new licensing structure allowing organizations to convert between user licenses and Flex Credits as business priorities shift, enabling more strategic allocation of AI investments.
- New Agentforce User Licenses and Add-ons: Unlimited employee-facing agent usage in a per-user-per-month (PUPM) model, with specialized offerings for Sales, Service, and Industries coming in Summer 2025.
While the Flex Credits and Flex Agreement are available immediately, the specialized Agentforce editions and add-ons are slated for release in Summer 2025, with “Flex Payment Models” launching in Fall 2025.
Moving Beyond the $2 Conversation Model
The pricing shift directly addresses the most significant barrier to Agentforce adoption identified in our previous research. The original $2 per conversation model drew extensive criticism from practitioners, with heavy adoption resistance particularly evident among mid-market organizations.
The consumption-based pricing created substantial budget uncertainty for organizations implementing Agentforce at scale. One support team leader we interviewed calculated that five agents handling 70 conversations daily would incur approximately $900 in daily costs – a prohibitive expense for many mid-tier organizations.
The $2 model created a paradox. Conversation-based pricing seems conceptually straightforward but practically resulted in unpredictable costs and adoption hesitation. Organizations were increasingly choosing competing overlaid AI solutions with more granular consumption metrics or fixed pricing.
Industry Reaction and Competitive Positioning
The new pricing model puts Salesforce more in line with competing enterprise AI platforms. Microsoft’s Copilot Studio offers both subscription and consumption-based pricing tiers, while Google’s Vertex AI Agents charges $12 per 1,000 queries, focusing on interaction volume rather than conversation sessions.
Salesforce is framing the change as a response to market demand for more flexible pricing that connects costs directly to business value.
“Salesforce’s new flexible pricing model allows us to use AI agents for different types of use cases far beyond customer service and traditional CRM and aligns our costs directly with the business value we achieve,” said Elia Wallen, CEO of Engine, in the official announcement.
Practitioner Perspectives: Moving in the Right Direction
Salesforce implementation specialists see the change as necessary but question whether it will be sufficient to drive widespread adoption.
“It’s a move in the right direction for sure,” says Andrew Russo, Salesforce Architect at BACA Systems. “The question is whether this will generate the momentum needed for the Salesforce community to start seriously implementing Agentforce, rather than just experimenting with it on demo projects.”
Chandler Ding, AVP Platform & Release Management at Toronto-based IGM Financial, highlights the broader governance challenges as Salesforce shifts toward consumption-based models: “As Salesforce expands consumption-based pricing across its portfolio, enterprises now face critical governance challenges. With multiple consumption points in products like Data Cloud and now Agentforce, we urgently need robust frameworks for capacity planning and forecasting. The technology is promising, but without predictable cost structures and comprehensive forecasting tools, many organizations will struggle to build business cases that satisfy finance teams.”
Russo’s and Ding’s sentiments echo what we have heard from many Salesforce professionals who have been hesitant to recommend Agentforce to clients, particularly mid-market organizations where budget predictability is critical.
While the new pricing model addresses the core cost structure issue, our analysis reveals a critical gap: robust forecasting and governance tools to help organizations manage their Flex Credits consumption.
The announcement mentions the Salesforce “Digital Wallet” with “detailed insights into usage trends, credit consumption, demand forecasting,” but lacks specifics on the capabilities. Organizations implementing Agentforce at scale will need:
- Proactive credit consumption alerts
- Department-level usage quotas and guardrails
- Anomaly detection for unexpected credit consumption
- AI agent performance optimization tools to minimize credit usage
Without these, organizations may still struggle to forecast and control costs, particularly during initial implementation phases when AI agent behaviors are being optimized.
Data Readiness: The Underlying Challenge
Our research indicates that pricing has been just one barrier to Agentforce adoption. The more fundamental challenge relates to data readiness within the typical Salesforce implementation.
Moving from $2 per conversation to $0.10 per action doesn’t solve the data quality problems that have plagued many early Agentforce implementations. Companies with fragmented, incomplete, or poorly structured Salesforce data still face significant implementation hurdles.
Salesforce implementations commonly suffer from inconsistent data practices, field utilization gaps, and cross-object relationship challenges that limit AI agents’ effectiveness regardless of pricing model.
A VE Economics Lens: Flex Credits and the Three Laws
Salesforce’s pivot to Flex Credits can be understood more deeply through the lens of VE Economics, which frames the business viability of Virtual Employees (VEs) using three foundational laws: Infinite Scale, Cognitive Commoditization, and Exponential Learning. Each law provides a unique diagnostic for assessing whether a pricing model supports, inhibits, or distorts the potential of AI labor.
1. Law of Infinite Scale
This law posits that AI labor, once created, should scale without traditional human capital constraints. The $2 per conversation model fundamentally violated this law by imposing a high marginal cost per interaction, effectively limiting the scale at which organizations could deploy agents. The Flex Credits model—charging $0.10 per discrete action—moves closer to honoring this law by allowing organizations to decompose work into micro-actions that can scale with greater financial efficiency. However, until credits can be purchased in smaller units or dynamically throttled, true infinite scale for SMBs remains aspirational.
2. Law of Cognitive Commoditization
AI agents should reduce the cost of performing knowledge work by breaking it into routinized, programmable units. Flex Credits align well with this principle by redefining “value” not as a vague conversation but as a quantifiable action—something that can be tracked, optimized, and priced transparently. We frequently hear the term “Jobs To Be Done” to define these actions. This redefinition is key to transforming AI labor from a bespoke consulting exercise into a standardized utility. However, without action-level attribution, firms may struggle to assign business value per Flex Credit used, thereby muting the commoditization benefits.
3. Law of Exponential Learning
The final law states that AI agents should improve over time without proportional increases in cost or complexity. The new pricing model could incentivize optimization of agent workflows and training data to reduce cost-per-outcome over time. Yet, Salesforce has not (yet) embedded this principle into pricing tiers—such as providing automatic credit discounts for high-performance agents or offering learning curve pricing ramps. Moreover, the absence of forecasting and governance tools hinders organizations from realizing compounding gains through reuse and refinement of AI agents.
In this light, Flex Credits represent progress, but not full compliance, with VE Economics. Salesforce has started moving from pricing AI agents like call center reps toward pricing them like cloud services. The next leap—perhaps in Agentforce 3.0—will be to embed learning efficiency and agent performance optimization directly into the economics of AI labor. Until then, the pricing may still feel like AI renting, not AI owning.
Progress, But More Work Ahead
The shift to Flex Credits represents significant progress in Salesforce’s AI monetization strategy. By aligning costs more directly with discrete value-creating actions rather than ambiguous “conversations,” Salesforce addresses a major adoption barrier.
However, mid-tier and SME Salesforce customers still face challenges:
- The $500 entry point for 100,000 credits may still represent a significant commitment for smaller organizations
- Lack of detailed forecasting tools creates uncertainty around budget planning
- The fundamental data readiness issues remain unaddressed
- Many implementation details around the Flex Agreement remain unclear
For enterprise customers with large Salesforce contracts and mature data governance, the new model likely removes a significant adoption barrier. For mid-market organizations – the backbone of the Salesforce ecosystem – the path to Agentforce implementation remains complex.
As Salesforce continues evolving its AI strategy, addressing these remaining challenges will be crucial to achieving the widespread adoption of Agentforce that the company clearly desires.
link