🎧 Prefer to listen?

I’ve been watching Salesforce announcements for years, and most of them blur together. This one didn’t. At Dreamforce 2026, Salesforce didn’t launch another app — it started dismantling its own interface, on purpose.

The headline product is a smarter Slackbot AI agent running on what Salesforce calls Slackforce, but the Slackbot is actually the least interesting part. The bigger move is AIforce: an umbrella over Claudeforce, Slackforce, and Agentforce Coworker — three products that all do the same thing in different places. They take Salesforce’s data, workflows, and business rules and deliver them inside platforms you already use. Claude. Slack. Microsoft Teams is next, and OpenAI after that. I covered the earlier agent comparison between these workplace suites in my Agentforce vs Copilot vs Google breakdown — this update changes that picture in a way that matters even if you’ve never touched Salesforce.

The interface is becoming optional — on purpose

Here’s the shift in one sentence: Salesforce is betting that you’ll stop logging into Salesforce.

The company now sells what it calls a Headless Toolkit — an architecture of MCP servers, APIs, plugins, and prebuilt skills that lets other applications operate Salesforce on your behalf. If you’ve read my explainer on what tool calling actually means, this is that concept deployed at enterprise scale. Instead of one chatbot inside one app, your CRM becomes infrastructure that Claude or Slack can query and act on directly.

Claudeforce is the clearest example. It’s a direct integration into Anthropic’s Claude with 37 ready-made sales skills — prospecting, pipeline cleanup, account research — all executed against your Salesforce data from inside a Claude conversation. Sales Cloud first; Tableau, Service, Marketing, and Commerce are on the roadmap. Slackforce works the same way in Slack, including Surfaces, which render live Salesforce dashboards you can filter and edit with your team without anyone opening a browser tab.

Agentforce Coworker sits inside Salesforce’s own interface but with the same ambition. It crossed 100,000 activated users in its first 35 days, and Adecco just rolled it out to all 27,000 employees. The Teams version could appear at any moment.

Why a no-code builder should care

You might be thinking this is enterprise news. It isn’t — it’s a preview of how every software subscription will work in about two years, and it lands on small teams first in one specific way: permission inheritance.

The part I find genuinely clever is that non-technical users can now build things on top of enterprise data without creating new security layers. Salesforce’s partners — AWS, Google, Lovable, Vercel — have already built integrations, so if your team gets permission to use Lovable, you can describe an app in plain English and have it query or even act on your Salesforce data. The app inherits your existing user permissions. No new access model, no consultant, no IT ticket. That’s the same pattern I covered in the governance layer solo builders need for AI agents — except here Salesforce built it for you.

If you run any kind of client business, this changes your build order. The CRM stops being a place you visit and becomes a service you call. Your “dashboard” is a Slack message. Your weekly report is a Claude conversation with your pipeline in context. I’ve argued before that your first AI workflow should start with tools you already pay for — Salesforce just made that argument for me, at enterprise scale.

The pricing mess is the tell

Now the honest part: nobody, including Salesforce’s own executives, can tell you what this costs.

The current stack includes per-seat cloud licenses, a $125/user/month Agentforce add-on, $5/user/month agent licenses, and Flex Credits at $500 per 100,000 actions — with a “standard” AI action costing 20 credits. Reporters asking executives how much a simple API call costs got no answer. Outcome-based pricing is apparently coming too, which would be a fourth model. Salesforce has publicly concluded that token-based pricing is too opaque, then shipped a credit system that is… also opaque.

This is worth watching because it rhymes with the wider mess I covered in the AI subscription price war. When the software layer disappears, per-seat pricing stops making sense — you’re not buying seats, you’re buying actions. Salesforce knows this; it just hasn’t figured out what to charge. My advice for now: don’t sign any AIforce-adjacent contract without a fixed-price pilot period, because the person selling it to you cannot tell you what a month of usage costs.

The strategic part everyone’s missing

There’s a reason Salesforce is giving away its own interface. Whoever owns the surface you work in — Slack, Teams, Claude — owns your attention. Salesforce decided it would rather power every surface than fight for one. It’s the same logic behind Microsoft and Google shoving agents into Teams and Gemini, and the same forced-adoption dynamics I described in the tokenmaxxing story about Amazon’s AI mandates — when the tool becomes invisible, usage becomes the only metric left.

For you, the practical read is this: the walls between your tools are coming down faster than the pricing models can keep up. That’s mostly good news. The MCP standard underneath all of this is the same pattern I explained in MCP vs skills — and it’s spreading from developer tooling into mainstream business software, which means your stack gets more capable without you migrating anything.

The bottom line

Salesforce’s Slackbot AI agent is the visible feature; the invisible story is a major SaaS company declaring its own interface optional and rebuilding itself as a layer other AI tools call. Small teams get real leverage from that — build on your CRM from Slack or Claude, with permissions already handled — but wait for pricing to shake out before committing budget. If you want a map of which AI tools actually fit a non-technical builder right now, start with the AI Tool Advisor or the start-here guide.