The Thesis That Waited A Decade

Most people think the AI era started two or three years ago. Emergence has been building toward it since 2016.

That year at our annual meeting, we presented a thesis we called Enterprise Behavioral Networks, the idea that machine learning would fundamentally invert the relationship between users and software. Cloud 1.0 made the user adapt to the software. Cloud 2.0 would flip that: software that watches how your best people work, learns what makes them exceptional, and coaches everyone else in real time. The end user, not the manager, not the CIO, would be the primary beneficiary.

We didn't have a name for it yet. By 2017, we did: Coaching Networks.

The thinking kept evolving from there. Why intellectual property is more than data in a spreadsheet; it's the way your people think and decide. Why knowledge walking out the door with departing talent is no longer inevitable. Why synthetic data isn't as safe as companies assume. Why model consolidation poses a real threat to the startup ecosystem. And ultimately, why only humans can break the rules that no longer serve us. AI optimizes, but humans mutate.

‍Conviction before consensus

This is how Emergence operates. We form a view early, before the market does. Sometimes before the technology is ready. 

When we backed Salesforce, most people thought enterprise software couldn't scale like consumer internet. When we became Zoom's first institutional investor, video-first work was still a niche idea. When we led Bland's Series B, 180 investors had already told the founder that voice was a dying medium and phone calls would be obsolete within a year. Bland has since handled over 175 million AI calls and just closed a Series C. In each case, the insight preceded the validation by years. The conviction had to hold through a long period where the evidence was thin and the skeptics were loud.

Coaching Networks was the same. In 2017, the infrastructure to do it well didn't exist. The models weren't capable enough. The enterprise data pipelines weren't in place. We published the thesis anyway, because the underlying logic was sound:

Humans are the only mutation engine in the age of AI.

Software can optimize. Only people can discover genuinely new moves. The job of the system is to find those breakthroughs and spread them to everyone else.

We were early. We knew it. And we kept working.

What we've always believed

The doom narrative is loud. “Half of white-collar jobs gone.” “Entry-level work eliminated.” A countdown clock on human relevance. We don't see it that way. 

Automation doesn't erase people. It moves the work. Every wave of technology in the last century has followed this pattern. It concentrates the highest-value tasks in human judgment, creativity, and the capacity to change the game itself. AI is not an exception. It's the most powerful version of a dynamic we've watched play out before.

What AI genuinely threatens is not the human role; it's the human advantage. If your people's best thinking flows into foundation models you don't control, your edge becomes part of the commons. What matters isn't "how much can AI automate?" It's "who owns the mutations?"

The technology finally caught up

Earlier this month, I published Above the Model, the most complete articulation of the Coaching Networks thesis to date. The core argument: the durable competitive layer in enterprise AI isn't the foundation model. It's the proprietary learning loop a company builds on top of it, seeded by its own people's judgment, compounding with every interaction. Generic intelligence commoditizes. The specific way your organization thinks and adapts does not.

A few days later, that argument went mainstream in a big way. The convergence was striking. When people thinking carefully about enterprise AI, from very different vantage points, land on the same core insight independently, it usually means the insight is right. And the attribution that followed was generous. What I'd offer in return is simply the chronology.

The reason these ideas feel newly obvious is that the technology has finally caught up to what they were always describing. That's the nature of being early. The thesis doesn't change. The world does.

What comes next

The opportunity in front of enterprise software builders right now is enormous. The models will continue to get better, and we’ll be watching the innovation eagerly. But from my perspective, the spotlight should be on the learning layer that compounds human expertise over time: domain-specific, organizationally rooted, structurally impossible to reconstruct from the outside.

Automate everything you can. Take every efficiency gain available. But don't let the optimization conversation crowd out the one that matters more. Keep the mutations yours.

Gordon Ritter is a Founder and General Partner at Emergence Capital. He has been developing the Coaching Networks thesis since 2016. You can reach him at gordon@emcap.com.