Below is a curated set of the highest-signal Salesforce + AI stories I tracked across July and August 2025 — the months when "Agentforce" stopped being a marketing word and started showing up in real production deployments.
Marc Benioff in the FT: AI won't decide our future — we will
The question isn't whether AI replaces us or amplifies us. It's how we design the partnership. Marc lays out the operating posture worth taking. Read the FT piece.
The four-concept literacy test for Salesforce AI
If you're leading a Salesforce deployment, you need to articulate the difference between LLMs, Agentforce, Data Cloud, and the prompt layer. For Agentforce you're combining all four — but you can swap LLMs without rebuilding the rest. Treat it like a stack, not a product.
You're wasting your time with agentic AI — if every agent lives in the same room
Not all agents belong in the same room — just like not all humans thrive in the same roles. A good post on mapping agent types onto org culture. Read it.
Start smart: how to train an Agentforce email agent like a human
Practical guide on getting started with Agentforce AI emails for service or sales. Read it.
RevOps tool consolidation: meet Bluebirds
If you're stacking Outreach, Gong, ZoomInfo, Chorus, and more — Bluebirds is the consolidation play to evaluate. Source.
57,000 unit tests, migrated in 9 months
What would have taken six years manually got done 8x faster with modular AI agents. The Salesforce Engineering case study of the year.
If you don't understand MCP, you don't know Agentforce
Full stop. MCP isn't a feature — it's the brain behind every intelligent decision. Memory, context, planning. Worth studying.
Slack + Salesforce: the underused goldmine
If you're using Slack and Salesforce together and you haven't wired Agentforce into the conversation, you're squandering it. Shockingly easy to set up.
McKinsey's GenAI-at-scale blueprint
I don't agree with every word — but the core insight is hard to ignore: one-off AI deployments don't cut it. Their study of 150+ orgs is required reading for any RevOps leader.
Stop running AI models all the time
Be smart about cost. Evaluate AI and non-AI options together. CRMA, Tableau Next, and the Salesforce Reporting layer cover a lot of ground without burning model tokens unnecessarily.
Workflow Rules are dead — and I'll miss them
The migration to Flows is real. Salesforce admins, this is the end of an era. The piece.
Real Agentforce use cases worth stealing
- Use case #1: Agentforce meets Service + Sales — leads pulled directly into the CRM
- Use case #2: Unstructured data, meet your agent — exploring across PDFs, emails, and notes
If you're brainstorming what to ship next, start with patterns that already work in someone else's org.
The Saurabh Gupta take on Agentforce for Service
Cost reduction is the obvious win. The hidden win: much faster resolution when the case is handed to the human-in-the-loop. The math compounds.