Are you a client?
Sign in to view the full news archive.
Salesforce has acquired London-based startup Convergence.ai, a company specialising in general purpose AI agents capable of performing complex, adaptive tasks in digital environments. The acquisition is expected to close in Q2 of Salesforce’s FY26.
Convergence was founded just a year ago by Marvin Purtorab and Andy Toulis, both veterans of Shopify and AI firm Cohere. The startup has quickly garnered attention for its technical sophistication and talent pool, which includes former employees from DeepMind, Meta, and PolyAI. While financial terms have not been disclosed, the startup has previously raised $12m in a pre-seed round led by Balderton Capital, with participation from Salesforce Ventures and Shopify Ventures. (See - Convergence raises $12m for personal AI agents)
Convergence AI’s solution is called Proxy, a generalist AI agent that doesn’t just process information, but will acquire new skills and improve over time. This is possible through a new type of AI model it is pioneering called Large Meta Learning Models (LMLMs). The big difference to other AI agent platforms is it integrates memory as a core component of the model architecture. Multiple instances of Proxy can all run at the same time, working across different departments and tasks, trained and directed by users.
Salesforce intends for Convergence to serve as the foundation of a new R&D centre in London, with the Convergence team playing a pivotal role in advancing Agentforce, Salesforce’s flagship AI agent platform, by contributing expertise in autonomous task execution and adaptive systems.
In parallel with the acquisition, Salesforce has introduced new pricing models for Agentforce. These include a consumption-based system using “Flex Credits”, where each AI agent action, such as updating a record or resolving a support case, costs 20 credits (equivalent to $0.10). A new “Flex Agreement” also allows organisations to shift spending dynamically between user licences and usage-based credits, with new simplified per-user pricing tiers also being released.
On the enterprise front, Salesforce unveiled a set of trust, security, and governance features designed to support secure deployment of AI agents at scale. These enhancements include integrated tools for data security, data governance in Data Cloud (featuring AI-based tagging and dynamic masking), and secure API management through MuleSoft. Agentforce also gains new safeguards, such as instruction adherence checks, sandbox testing with synthetic data, and a “Trust Layer” that moderates LLM interactions using data masking and toxicity filters.
Posted by: Simon Baxter at 09:38