Dreamforce 2026 turns agentic AI into an IT and CX integration job

Dreamforce 2026 is still weeks away, but Salesforce and the trade press covering the event are already telegraphing the real work: turning “agentic” from a keynote theme into integration, governance and service-ops change management.

Salesforce’s own “IT Leader’s Guide to Dreamforce 2026” frames the central tension as accelerating AI adoption without weakening data protection and compliance. UC Today, in an Aug. 4 guide by Sean Nolan, puts the same moment in operator terms: enterprises are moving beyond isolated copilots and pilots, and now have to connect agents, people and systems where work actually happens.

That combination is why Dreamforce is landing less as a “new feature” event and more as an architecture checkpoint for CRM platform teams, IAM/security owners, and contact center leadership who are now on the hook for outcomes.

Because of its scale, Dreamforce functions as a practical planning milestone rather than a marketing event.

Dreamforce 2026 runs Sept. 15, 17 at the Moscone Convention Center in San Francisco, with a free virtual program on Salesforce+ extending to Sept. 18, according to UC Today and Salesforce. Salesforce describes the conference as its 23rd year and says the program includes 50+ keynotes, 1,600+ sessions and 150+ hands-on learning experiences.

UC Today adds a point procurement teams often watch: Salesforce is listing a “Last Chance” full conference pass at $1,899, with a stated full price of $2,299, plus group registration options for teams of three or more. CMSWire’s event listing also reinforces the framing: Dreamforce brings together AI and CRM professionals to discuss AI agents, data solutions and CRM technologies, with a focus on the “agentic enterprise.”

Timing matters too. Network World’s Aug. 6 calendar puts Dreamforce directly alongside other enterprise-IT tentpoles in September, including Splunk .conf26 (Sept. 14, 17). For large IT organizations, that calendar density tends to force decisions about which teams attend which event, and which vendor roadmaps get validated first.

Headless 360 is a signal to start with APIs, policy and tooling

The most concrete pre-announced technical theme is “Headless 360.” Salesforce says it announced Headless 360 earlier in 2026 at TDX and is bringing it to Dreamforce as a core motif. In Salesforce’s description, “all of Salesforce” becomes accessible via API, “MCP,” or CLI so agents and humans can build on the same data, logic and governance already running the business.

For CIOs and enterprise architects, that is a procurement and operating-model signal. If Salesforce is serious about making its platform “headless,” future-proofing an Agentforce rollout becomes less about which UI flows look slickest and more about how quickly an enterprise can standardize integration patterns, establish guardrails, and operationalize change control across environments.

Salesforce’s IT guide spotlights keynotes explicitly organized around that shift, including a “Security Keynote” focused on trust across agents, data and platforms, and a “Headless 360” keynote framed around where humans and agents work. Even before any product launch at the show, the conference agenda itself is a hint about the implementation friction Salesforce expects customers to face.

CX leaders are being pushed to measure resolution, not containment

If Headless 360 is the platform story for IT, Agentforce is the operational story for service and experience teams. CX Today’s Aug. 18 guide by Sophie Wilson frames Dreamforce as Salesforce’s chance to demonstrate that Agentforce can improve real customer journeys, not just produce convincing conversations.

CX Today argues the test should be whether an AI agent can resolve concrete service tasks, billing issues, returns and entitlement checks, while preserving context across channels and knowing when to hand the case to a human. The piece also calls out a common measurement trap: treating “containment” or “deflection” as a proxy for actual resolution.

For operators, that suggests a simple shift in instrumentation. If an agent is allowed to complete tasks, service leaders need to track policy-compliant task completion, exception handling, and the quality and timing of handoffs, not just bot engagement rates. This is where data hygiene becomes operationally visible: duplicate profiles, missing order history and poorly maintained knowledge bases show up as longer handle times and more rework, even when the AI model itself is capable.

What to lock down before September: decisions that won’t fit in a breakout session

Dreamforce is built to be overwhelming, by design. Salesforce says there are 1,600+ sessions and 150+ hands-on learning experiences. UC Today adds 240+ community roundtables and a 400+ session virtual program. That volume is useful, but only if internal teams arrive with a short list of decisions they need to unblock.

Three practical threads connect the sources: Salesforce wants enterprises to build “agentic” capabilities on common governance, UC Today emphasizes cross-system connection as the premise of the Agentic Enterprise theme, and CX Today presses leaders to validate real-world resolution outcomes. Put together, Dreamforce becomes a place to validate whether the organization’s current integration and data posture can support autonomous action in production.

CMSWire’s listing offers another operational clue: the sponsor list includes global SIs and consultancies and large cloud providers, such as PwC, Boston Consulting Group, Amazon Web Services, IBM, KPMG and Slalom. That mix suggests many enterprises will be shopping for implementation capacity and operating-model help, not just licenses.

Questions to take into your Agentforce and Headless 360 evaluations

• Which workflows are being proposed for autonomous execution, and what is the approved policy boundary for each (refund limits, entitlement checks, identity verification)? Ask for a concrete task list, not a capability statement.
• What is the system-of-record path for the context an agent needs (customer profile, order history, knowledge, policies), and how is context persisted across channels and human handoffs? CX Today’s emphasis on cross-channel continuity is the failure mode to test.
• If Salesforce is pushing Headless 360 via API, MCP and CLI, what does the reference architecture look like for change control, secrets management, audit logs and environment promotion? The “headless” part moves the risk and the control points into platform engineering.
• What are the success metrics for service, and do they measure resolution quality, exception rates and rework, not just containment? Decide what will be instrumented in QA before the first pilot goes live.
• If the team is attending virtually via Salesforce+ (Sept. 15, 18), which hands-on sessions are essential to replicate in a lab environment after the show, and who owns that follow-through? UC Today notes virtual hands-on training is part of the program, but internal capacity determines whether it turns into deployable patterns.