Jensen Huang made something clear at Computex earlier this month: agentic AI isn’t coming, it’s here. “Useful AI has arrived,” he told the crowd in Taipei, describing a world where AI agents don’t just answer questions but observe, reason, plan, and act. It’s a seismic shift, and if you run a revenue organization, it should change how you think about competitive advantage.
Everyone right now is racing to feed their AI better data, which means more signals, more integrations, and better hygiene in the CRM. I understand the instinct and, quite frankly, it’s the obvious move. But, I think it’s the wrong race to win. The real differentiator is the intelligence layer purpose-built to prepare, structure, and route that data to other AI systems, not the data itself.
I’ve spent the better part of my career watching the same problem repeat itself in sales technology. The challenge has always been that CRM runs on self-reported information and data that reps enter manually, selectively, and inconsistently. Everything downstream inherits that noise. When we started applying AI to this problem, we were solving for something harder: how do you take real activity (emails, calls, meetings, the messy reality of complex enterprise selling, etc) and transform it into answers that are actually trustworthy?
That work is not fast because it’s not something you can recreate by pointing a LLM at a raw data lake and hoping for the best. Take a company like Microsoft selling to Verizon, for example. You’ve got dozens of sellers, specialist overlays, multiple stakeholders across different business units, and deals that span years. When a CRO sits down with a CFO, they might cover fifteen different topics in an hour. Matching that activity to the right context whether that’s an account, opportunity or moment in the sales cycle is an extraordinarily difficult problem. We’ve been solving it for a decade, across billions of transactions, and there is simply no shortcut for that institutional knowledge.
This is why I believe the companies that win the AI era will be the ones who’ve built AI that makes their other AI smarter. Competitors can copy a data strategy, license the same signals, build the same connectors, and chase the same integrations. But what they can’t easily replicate is a bespoke reasoning and enrichment layer trained on years of real sales context. That’s a defensible architectural advantage and it becomes more durable the more it learns.
What I find compelling about the moment Jensen described this week is how it reframes the entire conversation. In an agentic world, your AI is serving other AI systems, not just humans. The prep work is no longer background infrastructure, but becomes the core product. Customers need the right answer, right now, grounded in what’s actually happening in their business, delivered wherever an agent or a human happens to be working.
Here’s what I don’t believe, and it matters: I don’t believe human leadership disappears in this model. Instead, what changes is where leaders spend their time. Today, too much of a sales manager’s day is consumed by gathering pipeline updates, chasing forecast calls, synthesizing rep activity into something a VP can act on. AI can coordinate and surface that information directly, freeing up managers to do what only humans do well: build relationships, read the room, make judgment calls, and lead.
The manager of the future is operating at a higher altitude because AI is handling the information processing below them. The companies building toward that future understand something that the pure data-play vendors don’t: raw data is a commodity. Intelligence that can be trusted by humans and by the agents increasingly acting on their behalf is the moat. And that moat is built through years of hard, unglamorous work on the problem beneath the problem.
We’ve been doing that work for a long time. The market is finally ready to see why it matters.
Jason Ambrose is CEO of Backstory (formerly People.ai), the leading AI data platform for revenue teams.
