What Salesforce announced
Dreamforce 2025 centred on the "Agentic Enterprise": Agentforce 360, deeper Data Cloud integration, and a $1.5bn commitment to San Francisco tied to the agent platform build-out. The headline was scale. The substance was a coherent stack for building, governing, and running AI agents against enterprise data.
Why this matters beyond the Salesforce ecosystem
Every major platform vendor is now shipping an agent layer: Microsoft Copilot Studio, Google Agentspace, ServiceNow AI Agents, Workday AI, and independent frameworks from OpenAI, Anthropic, and LangChain. The strategic question is no longer "should we adopt AI agents?" but "against which data, with what governance, and integrated where?".
The three shifts leaders should plan for
1. Agents become the default UI for enterprise software
Users increasingly interact with systems through a conversational or agent-driven surface rather than by navigating menus. This changes UX design, training, and adoption planning.
2. Data foundations become the deciding factor
Agents are only as useful as the data they can trust. Enterprises with clean, joined-up data (a company knowledge graph) will scale agents fast. Those without will keep piloting.
3. Governance becomes a board-level concern
When software takes actions autonomously, ownership, audit, and accountability move up the priority list. Expect the CFO, GC, and CISO to be equal stakeholders alongside the CIO.
What to do in the next 90 days
- Audit your data foundation. Can an agent get a trustworthy answer to your top ten operational questions today? If not, that is the first investment.
- Pick two bounded workflows. One customer-facing, one internal. Deploy agents against them with clear KPIs.
- Establish an agent governance policy. Ownership, review cadence, escalation, audit trail. Sign off at ExCo.
- Upskill your leadership team. Not on the vendor stack, on how to make decisions about autonomous software.
- Instrument outcomes. Cycle time, resolution rate, cost per transaction, customer satisfaction. Weekly review for the first quarter.
The healthy scepticism
Vendor keynotes overstate maturity. Real deployments in 2025 still hit friction around data quality, integration, and change management. That is not a reason to wait, it is a reason to invest in the foundations that make agents earn their keep. See our view on the knowledge graph foundation and how to build it inside our three-step method.