Five Things (Other than “Headless 360”) That Actually Matter From Salesforce TDX 2026

I just returned home from TDX with my usual mix of energy and skepticism that any good conference produces. There was a lot of big-stage excitement —  Golden Hoodies! New products! Live demos that almost worked! — but underneath all of it, a few things stuck with me that I think matter beyond the hype cycle.

Here's what I'm still thinking about.

1 - Slack Isn't a Messaging App Anymore

This is the one I keep coming back to.  

About five years ago – before Salesforce even bought Slack – the real power of Slack clicked for me when I saw it used correctly. Slack isn’t just a communication tool, but as I used to preach to anyone who would listen, its real value is as a control plane for your business.  A Slack integration subsequently became “table stakes” for any app I considered from my business, and I’ve never regretted that course. 

At TDX, Salesforce presented Slack as the "Agentic Work OS" and the "front door to Agentforce 360." That's marketing language, sure, but what they actually demonstrated was something more substantive: Slack is being repositioned as the primary interface through which people will interact with agents — not just talk to each other.

Think about what that means practically. Right now, most organizations use Slack (or Teams) as a communication layer that sits next to their CRM, their ticketing systems, and their data. Agents live somewhere else. You log into a portal, or you go to a specific tool.

What Salesforce is building is a world where the agent comes to you — in the channel where the work is already happening, with context about what's being discussed, able to take action rather than just answer questions. A 47% productivity increase was cited for organizations using Slack-native agents, which I take with all appropriate salt, but the directional point is real.

This especially matters for workers who are early in their journey with agents.  We all learned how to “google” by adopting a very terse language.  You dropped the filler words and entered only the key words for the pages you needed. This approach is exactly the opposite of what works best with Agents, which thrive on context. Effective use means getting conversational, and by positioning agents in Slack you’re not only providing that context but you’re cueing users to shift their framework and interact with an agent how agents perform best – conversationally.  

For clients wondering where to focus AI adoption attention: your employees already live in Slack. That's where agents need to live too.

2 - The AgentExchange Creates a New Job Description

Salesforce announced a new unified marketplace — AgentExchange — combining what used to be the Slack Marketplace, AppExchange, and a new agent-specific catalog. There are already 60+ curated agents available, and they backed it with a $50 million fund for agent builders.

Here's the part that didn't make it into the keynote slides: someone has to curate this.

Installing an agent isn't like installing an app. Apps do a defined thing. Agents make decisions, take actions, interact with data, and operate with some degree of autonomy. The gap between "we installed the procurement agent" and "the procurement agent is working well for our organization" is enormous — and it requires ongoing attention, not a one-time implementation.

This is a new role. Not a new title necessarily, but a new function: the person or team responsible for deciding which agents your organization uses, how they're configured, what data they can access, how their performance is monitored, and when they should be retired or replaced.

In the Salesforce ecosystem, that responsibility is going to fall somewhere. In my experience, it falls to the consultants who helped build the thing — which means this is also a new service offering, and one that clients should start asking about now.

3 - Your Knowledge Problem Is a Data Problem in Disguise

The hands-on session I attended on Enterprise Knowledge was, honestly, one of the most practically useful things I've done at a conference in years.

Here's the distinction that matters: Traditional Salesforce Knowledge requires you to manually create articles, migrate content, and maintain a curated library inside Salesforce. Enterprise Knowledge connects to where your content already lives — SharePoint, Confluence, your website, Box, YouTube — and uses semantic search to make that content retrievable by agents.

"Semantic search" sounds fancy, but the practical implication is simple: instead of an agent finding only exact keyword matches, it finds content based on meaning. Ask an agent "what's our return policy for damaged goods?" and it retrieves the right policy document even if it's never used those exact words.

The stat from the governance session that's been rattling around in my head: 60% of organizations say AI is critical to their strategy, but only 12% feel their data is ready to support it.

That gap is the whole ballgame. You can have the best agents in the world and they will confidently give wrong answers if they're not grounded in clean, current, accessible data. Enterprise Knowledge is Salesforce's answer to the grounding problem — and it's the path forward for any org that wants agents that are actually trustworthy, not just impressive in a demo.

4 - The Empathy Gap Is Real, and It's a Competitive Advantage

This one was harder to capture from slides because it showed up more in the framing of sessions than in any single announcement.

The leadership content at TDX kept circling back to a theme: in an AI-heavy environment, the differentiator for technical leaders isn't technical depth — it's human connection. The ability to bring people along. To read a room. To build trust in a change process that, let's be honest, feels threatening to a lot of people.

Sarah Franklin's return to the Salesforce stage was a moment — Patrick Stokes credited her with building community over a decade, with "pulling people in." That's not a product feature. That's a leadership quality, and it was being held up as aspirational on purpose.

For consultants and technical leaders, I think this translates directly: the clients who succeed with AI won't be the ones with the best tech stack. They'll be the ones whose leaders can explain what's changing, make people feel included in the transition, and build genuine enthusiasm rather than compliance.

That's a skill worth developing deliberately… and it’s one that’s near and dear to our hearts at BrightHelm. In practice, it looks like a commitment to meeting people where they’re at, making sure you’re building a big enough boat, embracing vulnerability, and leaving room for growth.  

5 - The Real Metric Is Adoption, Not Deployment

This was the undercurrent running through almost every session, and I want to name it explicitly because it often gets lost in the excitement of building things: Success is adoption, not deployment. 

In nearly every session I attended, albeit in different ways, success was (wisely) measured by whether people actually use what's been built — not whether it was deployed on time and under budget.

Salesforce's own internal example was instructive: they powered their 1-800 support line with Agentforce voice, and they told the room to call it. Not "here's a case study," but "go try it right now." Their help site (help.salesforce.com) is built on the same platform. They're eating their own cooking and measuring the experience, not just the implementation. And sure – that’s been a rough road.  Far from perfect, or in some cases, even acceptable. But they have had the guts to implement their own product and take the feedback.  

More importantly, this means the conversation about AI investment shouldn't end at go-live. The question worth asking before you build anything is: what does adoption look like in six months? Who's using it, how often, and what happens when they don't?

The organizations that answer those questions before they deploy are the ones who will actually see returns.

The Bottom Line

TDX 2026 was, in many ways, a coming-out party for the agentic enterprise as a real, tangible thing rather than a concept. The demos were more polished. The customer stories were more credible. The tooling — AgentScript, the testing frameworks, the observability layers — showed a platform that's starting to mature around the hard problems.

But maturity in the platform doesn't mean maturity in how organizations use it. The gap between what's possible and what's actually working in the field remains wide — and that gap is where good consulting lives.

The five things above aren't just interesting product developments. They're signals about where the real work is, and where the real value is going to be created over the next few years.