THREAD
2 min read · December 2025

5 things every MarTech leader gets wrong about Agentforce

A short list, written after my third "wait, Agentforce can't actually do that" conversation this month.

AgentforceSalesforceAI

I've had this conversation three times this month. Time to write it down.

1. "Agentforce will replace our service team"

No. Agentforce augments, it doesn't replace. It handles tier-1 queries and routes complex issues. Your service team still matters — they're just working on harder problems.

2. "It works out of the box"

Also no. Agentforce needs training data, defined topics, tested flows, and ongoing refinement. Budget 3-6 months for a production-ready deployment, minimum.

3. "The AI just figures it out"

It doesn't. You need to explicitly define what topics the agent can handle, what data it can access, and what actions it can take. The "AI" is smart within the guardrails you set.

4. "We can launch without IT"

You can start without IT. You cannot scale without IT. Data access, security reviews, integration with existing systems — you'll need technical partnership.

5. "ROI is immediate"

The ROI is real, but not immediate. Plan for 6-12 months before you see meaningful metrics movement. The first 90 days are learning, not earning.

The Bottom Line

Agentforce is genuinely useful. But it's useful in the way a CRM is useful — when implemented thoughtfully, with clear use cases, and realistic expectations.

Don't believe the demo. Plan for the work.