Dreamforce 2025: Benioff's AI Push Meets a Slower Market Reality
At Dreamforce 2025, Marc Benioff declared the rise of the "agentic enterprise." AI, he said, now sits at the core of Salesforce. Yet he admitted what most leaders feel privately: adoption is still slow across large companies.
"Customers are getting their head around how to deploy AI," he said. The pace of product innovation is outpacing customer rollouts by a wide margin.
Agentforce Is the Platform - But Many Companies Aren't Ready
Salesforce doubled down on its AI vision even as investor patience thinned. Shares are down 34% from the December 2024 peak, and the company cut 4,000 support roles after pushing AI agents deeper into operations.
Benioff's stance: the market hasn't fully grasped how central Agentforce has become. "People don't understand that Agentforce is part and parcel of Salesforce. It is the core of every product we make now; it is the platform."
Human + AI: A Realistic Productivity Model
Anthropic CEO Dario Amodei pushed back on the idea of a 100% AI workforce. He expects AI to handle the bulk of code in the near term, but not all of it. "In six months, 90% of code would be written by AI models. They [human engineers] can focus on the 10% that's editing the code or writing the 10% that's the hardest, or supervising a group of AI models… you just end up being 10 times more productive."
It's a healthy framing for leaders: automate the repetitive, keep people on the needle-moving work. Supervision and exception handling still matter.
The Real Blocker: Architecture, Not Appetite
Benioff called out the main constraint: legacy systems. "These customers have to go back and modify massive architectures they have and systems they're running."
There are bright spots. Williams Sonoma and Pandora were cited as companies that have moved ahead with Salesforce's AI products. But for most enterprises, AI rollouts are still early-stage.
Vision: Be "Customer Zero" and Show the Path
Salesforce plans to prove the model internally, then publish the blueprint. "What we realized at Salesforce is we have to be customer zero… we have to be able to go from one, to two, to three, to four, to five and show you what we're doing."
Martin Schneider, VP & Principal Analyst at Constellation Research, supported the direction, noting it "makes a lot of sense when you consider how much vibe coding, and AI in general, is going to further commoditize the CRM application/pricing model."
What's New: Product Moves You Should Track
- Agentforce 360 platform - Salesforce's broader AI agent layer across products (official page).
- Service Cloud renamed to Agentforce Service - signaling AI-first service operations.
- New ITSM product - a clear step into enterprise service management with AI agents at the center.
For Management: A Practical Action Plan
- Set a target use case, not a moonshot. Pick one agent workflow (support triage, lead routing, code review) and ship a pilot in 60-90 days.
- Audit data and systems. Identify the top three blockers: identity, permissions, data quality. Fix those first.
- Define "human-in-the-loop." Spell out when agents decide, when they suggest, and when people must approve.
- Measure the right things. Track cost per ticket, time-to-resolution, first-contact resolution, and customer NPS-not vanity metrics.
- Stand up a small "Agent Ops" function. Own prompts, policies, escalation rules, and model updates.
- Reskill your team. Shift roles toward supervision, quality control, and exception handling. Train for it, don't hope for it.
- Modernize in slices. Strangle legacy dependencies one process at a time instead of big-bang replacements.
- Communicate the new org model. Be explicit about job changes, new KPIs, and how success will be rewarded.
The Bottom Line
Salesforce is all-in on agents. The tech is ready enough for practical wins, but most enterprises need systems work, new roles, and clear guardrails to see real value.
If you're leading this shift, start small, prove ROI, and build the muscle. The companies that productize their own AI playbooks the fastest will set the pace for everyone else.
Related resources: Anthropic on AI systems and safety; Salesforce's Agentforce overview above for product specifics. If you need a structured path to upskill teams by role, explore AI courses by job.
Your membership also unlocks:
