VENDORiQ: Salesforce’s New AI Pricing Model Delivers Transparency in an Increasingly Desperate AI Cost War

Why it Matters

This announcement reflects a broader trend in the technology industry where vendors are seeking to rationalise the cost of AI services. Many providers initially offered AI at narrow margins, or even at a loss, to gain market share. However, rising costs of increased consumption due to complex, multi-AI call workflows and the increasing viability of open-source generative AI models have put pressure on these initial pricing strategies. 

For example, OpenAI’s strategy involves managing AI consumption through dynamic rate limits on its API, which function as effective usage caps based on requests per minute (RPM), tokens per minute (TPM), and other metrics. These limits vary by model and the user’s tiered access, which is typically unlocked by cumulative spending. Pricing is granular, based on a per-token fee for language models and a per-image fee for image generation. For pro products like ChatGPT Plus, message caps are often implemented but with less explicit transparency, leading to ‘dynamic’ usage limits.

Google’s Gemini Apps, including those integrated into Workspace, also feature ‘usage limits’ designed to manage capacity. This means they may cap the number of prompts and conversations within a specific timeframe, with capacity replenishing dynamically. Higher limits are typically provided to users on paid tiers, such as Google AI Pro or Ultra plans. Some Google Workspace plans with Gemini features also have explicit monthly usage limits. Both OpenAI and Google, similar to Salesforce, are navigating the challenge of fairly monetising complex AI interactions sustainably while prioritising security and governance.

Salesforce’s move with Flex Credits appears to be a direct response to these evolving market dynamics. It offers a far more transparent approach to AI costs than other vendors, suggesting a way forward.

The shift to a consumption-based model for AI actions, rather than conversations or broad subscriptions, aims to provide greater transparency and control over AI spending. While this granular approach can be beneficial for procurement and finance teams seeking to align expenditure with actual usage (demand fulfilment), it also introduces a new layer of complexity in forecasting and budgeting, which will need careful demand management and enterprise cost controls. 

Salesforce customers can monitor consumption via the Salesforce Digital Wallet to understand usage trends and predict future credit needs. The ability to convert user licences into Flex Credits and vice versa, as part of the Flex Agreement, offers a degree of financial agility that could prove valuable as AI adoption matures and strategic priorities shift.

Clients of other AI services will need to follow similar precautions.

However, the most important takeaway is that AI costs will likely increase sharply over the next three years, in part due to vendors being unable to continue with ‘loss-leader’ practices, and with growing adoption of AI orchestration. 

Who’s Impacted?

• Procurement Teams: Will need to understand the new consumption-based pricing models of AI, work with business owners to forecast usage and assign business value to validate supplier value creation outcomes.
• CIOs/CISOs Will need to assess the financial implications of the new pricing models and be prepared to communicate the likely budget increases required.
• Business Owners: Will need to align costs with benefits/realisation of their objectives within budget constraints. 

Next Steps

• Conduct a detailed cost analysis: Procurement, AI teams and business owners should work together to model potential costs of changing AI system, considering various use cases and expected agent action volumes.
• Engage with Salesforce: Seek detailed clarification on specific credit consumption rates for complex agent actions and explore the full potential of the Flex Agreement for dynamic resource allocation.
• Implement appropriate demand management: Work with functional users to align supply, demand, costs, and benefits – monitor and regularly review/update controls to maximise value capture outcomes.