Date: 13 Oct 2025 (Dreamforce runs 14–16 Oct 2025 in San Francisco)
When Salesforce is life and AI‑powered productivity is the new oxygen, even your Slackbot needs an agent.
Dreamforce always kicks off with big themes. Last year it was generative AI; this year it’s the Agentic Enterprise. Instead of replacing humans, Salesforce’s vision is to put software agents at your side to automate the rote stuff and leave you to do the thinking. So far there’s no giant robot on stage, but there are dozens of brand‑new product names and at least one philosophical debate about whether we should trust a bot with a sales quota. And yes, that means another round of rebrands to memorize — because what’s a Salesforce event without a fresh wave of naming confusion to keep our customers guessing? Here’s the quick‑and‑dirty digest of what hit the wire on the eve of Dreamforce 2025.
☁️ Agentic Enterprise: humans + AI side‑by‑side
Salesforce’s core announcement is the Agentic Enterprise. Think of it as a framework where humans collaborate with autonomous software agents instead of being replaced by them. In practice that means an entire stack – Agentforce 360 for building and running AI agents, Data 360 for feeding them clean data, Customer 360 apps (now rebranded as Agentforce Sales, Marketing, Service, etc.) and Slack as the conversational interface.
According to Salesforce, AI agents never sleep, never drop a lead and never keep customers waiting. They’re there to augment humans. Still, one can’t help but wonder — will this really empower humans or quietly substitute them over time? That’s the ethical question hanging over every demo. Marc Benioff can wax philosophical all he wants, but the real story for admins and devs is the tooling that makes these agents possible — or that will replace them… but I’m not here to get depressed.
🔧 Agentforce 360 platform
The Agentforce 360 platform (yep, another rebrand) sits at the center. It’s the toolset for building, running and monitoring AI agents. Big thanks to SalesforceBen for all the details shared in preview!
New features include:
| Feature Quick takeaway | |
|---|---|
| Agentforce Builder | A conversational dev studio that lets you describe an agent in natural language, automatically generating workflows and code. No need to write LLM prompts by hand — the builder guides you through tasks, policies and integration points. |
| Agentforce Voice | Native voice capabilities for IVR and phone bots. Think call‑center agents that speak like a human, pull context from CRM and route to the right person. |
| Hybrid Reasoning + Agent Script | Salesforce is blending deterministic flows with large‑language‑model reasoning. Agents can follow step‑by‑step scripts for predictable tasks but can also reason with unstructured data when the unexpected happens. You can write custom logic using Agent Script (a new low‑code domain‑specific language). |
| Agentforce Vibes | Because even AI agents need a vibe check. This low‑code tool lets you refine an agent’s tone and personality. It’s part of the push to make agents feel less like a bot and more like a coworker. |
| Observability | Built‑in dashboards and logs so you can see what your agents are doing, measure success and troubleshoot. |
| Intelligent Context | Uses Data 360 to extract structured data from emails, PDFs and other unstructured sources and feed it to agents automatically. |
In plain English: Salesforce knows that building an effective AI agent is more than gluing ChatGPT to a Slack channel. They’re shipping a platform to orchestrate data, logic, voice, personality and monitoring – and they want you to use it for everything from sales follow‑ups to IT tickets.
💬 Slack as the Agentic OS
Slack officially becomes the Agentic operating system. It’s no longer just a chat app; it’s where humans and agents talk in real time. Key bits:
- Slack‑first apps – Salesforce is rebuilding its apps (Sales, IT Service, HR Service and the analytics tool Tableau Next) to run inside Slack. That means you can create pipeline records or close an incident without leaving chat.
- Channel Expert Agent – A new AI agent lives in each Slack channel, summarising threads and prompting you for next steps. It uses the same Agentforce engine, but with Slack context baked in.
- Enterprise Search – Ask Slack to search across Google Drive, GitHub, Dropbox and your CRM, and it returns answers (not just files). It’s basically ChatGPT over your enterprise data, with Salesforce promising rigorous permissions.
- Reimagined Slackbot – The OG Slackbot becomes a personal agent that can schedule meetings, answer questions and trigger automations. There’s also an API (
Real‑Time Search) and a Model Context Protocol to plug in external LLMs like Anthropic, OpenAI or Meta. - Pricing tiers – Salesforce teased new paid Slack tiers (Pro, Business+ and Enterprise+) with these agentic features rolling out through 2026 — ouch, it’ll probably hurt!
Behind the jargon is a simple reality: Slack wants to be the place where you do things, not just talk about doing them. If your daily workflow lives in Slack, embedding AI agents there makes sense.
But the real question is: are companies actually ready to abandon their trusted, working tools for an AI‑powered Slack — swept up in the AI frenzy — or will they just keep paying for what already works?
📊 Data 360 & Customer 360 apps
Salesforce’s data platform (formerly Data Cloud) is now Data 360, emphasising unified profiles and context extraction. It feeds the agents with structured, governed data. And yes, remembering that the Customer Data Platform became Data Cloud was already a headache for customers — now with Data 360 in the mix, good luck explaining that DC360 in emails isn’t a new postal code. At the same time, the Customer 360 apps are rebranded as Agentforce Sales, Agentforce Marketing, Agentforce Service and so on. The goal: every part of your CRM becomes agent‑powered.
🎧 Agentforce IT Service: AI helpdesk in Slack
Another shiny new product is Agentforce IT Service. Salesforce positions it as a conversational‑first IT service desk built natively on Salesforce. You spin up multiple AI agents that know different domains (e.g., password resets, laptop ordering). They live in Slack, Teams or your web portal and can answer employee questions 24/7.
The product includes:
- Agentic service desk – The interface where IT can build and manage AI agents and track performance.
- Multiple AI experts – Agents are trained on knowledge bases, system logs and existing tickets. They can troubleshoot automatically and hand off to humans when needed.
- Configuration Management Database (CMDB) – A new agentic CMDB uses a service graph to map systems and dependencies, helping with root‑cause analysis.
- Automated incident management – The platform detects issues, opens incidents and even remediates them where possible.
General availability for the complete solution is October 2025 – yes, just in time for your 2026 IT budget.
🕸 MuleSoft Agent Fabric: orchestrating the agent sprawl
Everyone is building agents, but connecting them can be messy. Enter MuleSoft Agent Fabric, a new solution aimed at preventing “agent sprawl.” Salesforce pitches it as a single place to discover, orchestrate and govern agents across your enterprise. Four main components:
- Agent Registry – A catalog to register and discover any agent or tool, whether built with Agentforce or a third party
- Agent Broker – Orchestrates and coordinates agents across ecosystems so they can call each other without creating an infinite loop
- Agent Governance – Enforces policies and security controls, ensuring your agents don’t leak sensitive data or go rogue
- Agent Visualizer – Provides real‑time insights into how agents are performing and interacting
This might sound like overkill, but if every team starts spawning agents, having a control tower makes sense.
Funny thing is, I can’t even spot the “360” in the product name — ironic, considering it’s supposed to tie everything together.
🧠 Final thoughts (from a nerd @ work)
Dreamforce 2025 is shaping up to be a turning point for Salesforce’s product stack. The company has done what it always does: rename a bunch of things, introduce a new layer of abstraction and promise that AI will make everyone more productive. Cynicism aside, there’s a clear through‑line:
- Data (Data 360) is the fuel
- Agents (Agentforce 360 and its siblings) are the workers
- Slack is the interface
- Governance (MuleSoft Agent Fabric) keeps it all from turning into a dystopian bot war
The Agentic Enterprise vision might sound like buzzword bingo, but the pieces here are tangible. If you already build flows and automations in Salesforce, the new builder and script tools could save hours. If your users live in Slack, the new Slackbot and Channel Expert will turn chat into an action hub. And if your organisation has dozens of AI initiatives, an agent registry and broker might prevent chWhat’s missing? Pricing details, real‑world limits and a plan for how to manage ethics when agents start making decisions. Also, we still need to see these demos survive beyond the Moscone main stage. But for now, grab some popcorn and tune in: Dreamforce is officially started. And if you’re reading this from Europe (like yours truly), just remember that San Francisco is nine hours behind – the keynote might double as a late‑night show.
Don’t miss the main keynote where Marc Beniof will probably reveal something else.
1 commento