Salesforce advocates for a shift in enterprise AI from raw model power to context-driven systems, highlighting new tools like Data 360 and the Agent Context Engine that enhance relevance, memory, and governance for improved customer service and operational efficiency.
Salesforce is making a broader argument about enterprise AI: that the quality of an agent depends less on raw model power than on the business context surrounding it. In a blog tied to Dreamforce, the company said an AI agent can easily make a sensible recommendation from purchase history alone and still fail badly if it misses a live service issue, such as a complaint that has not yet been resolved. The point is straightforward. Relevance is not enough if timing, history and permissions are absent.
At the centre of Salesforce’s pitch is Data 360, which the company says can unify customer profiles, preserve consent preferences and let teams build audiences from plain-language requests rather than long chains of filters. That idea is meant to reduce the manual work of campaign set-up while improving the accuracy of targeting. A related Salesforce white paper with partners on Data 360 and MuleSoft argues that context is what gives data meaning, history and relationships, which in turn improves AI performance and supports more proactive service.
Salesforce also highlighted the Agent Context Engine, which is designed to combine structured records with unstructured material such as call transcripts, policy documents and internal knowledge articles. The company says this allows an agent to answer questions using the information it is permitted to access, while memory can carry approved facts from one interaction to the next. In practical terms, that is an attempt to solve one of the most common weaknesses in enterprise AI: systems that can retrieve facts, but do not understand which facts matter in a given business process.
Another part of the argument is employee productivity. Salesforce said Agentforce Coworker brings business information and actions into a single conversational interface across Salesforce, Gmail and other enterprise applications. The company says this is intended to reduce time spent searching across systems before meetings or customer calls. It also points to Data 360 Headless and the broader AIforce approach, which is meant to extend trusted business context into other surfaces, including Slack and Claude, rather than keeping it locked inside Salesforce.
The Dreamforce keynote framed the challenge in five parts: data, efficient access, semantics, memory and governance. That is a useful checklist for companies testing AI in customer-facing work. A recommendation engine, for example, can appear intelligent while still overlooking an open support case, a recent policy change or a restricted field. External case studies on customer personalisation and service automation suggest that when systems do combine purchase history, policy rules and channel context, they can improve retention, satisfaction and resolution rates. Salesforce’s message is that similar gains are possible only if the underlying context is current, connected and controlled from the start.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
