Lately I’ve been reading a lot about trust and control in agentic systems, and I keep feeling an old tension come back into focus. The more I sit with the current wave of AI discourse, the more I find myself turning to an old friend—one I treated with a lot of suspicion earlier in my life: governance.
I have had mixed experiences with governance. I’ve seen it as category creation, compliance burden, and quiet enabling structure. I’ve run into the problem at Systinet, State Street, Pivotal, VMware, and now again in the age of AI and agents. Underneath all of those experiences was the same recurring tension. There’s a pull between new ways of working and old ways of understanding.
That tension has always excited me. I have strong anti-authority instincts coupled with a (potentially stronger) attraction to structure and systems. I appreciate constraints that shape progress, and hate the ones that get in the way. I’m made for the froth where new ways of working collide with old instincts.
I first met my old friend at Systinet. We were trying to build momentum around one of the most over-engineered and least-loved parts of the radically over-engineered SOA stack. We built that momentum around the idea of governance. My team worked with customers to get their implementations moving, which meant making something pretty unexciting matter to people who already had mixed feelings about it before we ever walked in the room. Governance was the kind of word that made people brace themselves for friction, overhead, and somebody else’s checklist. That’s where I first learned to separate useful governance from deadening control.
I developed a framing around three pillars: visibility, trust, and control. I’ve turned to it in every job I’ve had since. The technology landscape changed, but the language still held. It held because it gave me a way to talk about governance that made room for safety and speed at the same time, without pretending one had to cancel the other out. It helped me think about the constraints of regulated environments. I used it to steady nervous leaders in the cloud-native era. Now it helps me understand how AI and agents challenge my own assumptions about what’s safe and what’s not.
Visibility tells us what we have, how it’s working, and where that varies from what we expect. Trust tells us where we can safely lean on people, systems, and agents without turning every decision into a brawl. Control helps us intervene in the lightest possible ways without turning loose our fears with a badge and a gun. Getting that balance right is what makes governance an enabler rather than a tax.
Coding agents and agentic employees help us move faster than we ever have before. We’re also spending our days working with systems that are harder to fully understand because they unsettle the core ideas we were trained in. These systems let us build things we would have left fallow before. They also fractal the amount of software and decision-making we are responsible for. The result is more systems, more surface area, and more consequential decisions happening out of our direct line of sight. This is exactly the kind of moment that calls governance back into the room.
It’s a natural response to call for governance and ask it to save us. The question is how we want to do that. We can opt for the theater of corporate control that helps anxious institutions sleep at night. Or we can create enough visibility and trust to establish a pragmatic discipline that harnesses speed without ignoring its risks. One path stifles the moment. The other helps us thrive in it.
I’ve often thought about governance and compliance the way I think about taxes. It’s a trope that nobody likes paying, and that’s true in the abstract. Yet people accept them when the benefits are clear, visible, and close enough to feel. The same is true here. Governance feels like a tax when it obscures the value it creates. It becomes an enabler when the visibility, trust, and control it buys are real enough for people to feel in their day-to-day work.
How do we frame those tradeoffs? The best tool I’ve found is the one I’ve carried for decades: visibility, trust, and control.