Field notes on runtime governance.
Perspectives from the team on deploying, governing, and measuring autonomous AI agents in the enterprise.
Policy on Paper
Your AI policy says your agents cannot access customer data without authorization. Your agents can, in fact, do exactly that. Both statements are true. That is the problem.
April 20, 2026The Autonomy Paradox: How PE's Control Instinct Destroys the Value It Paid For
PE firms acquire companies and immediately tighten controls: standardized reporting, mandated tooling, new approval workflows. From the boardroom, it's responsible governance. From the engineering floor, it's the moment top performers update their LinkedIn. Fifty years of research and the DORA data say the same thing: this playbook destroys the value the deal was meant to create. Here's the boundary model that resolves the paradox.
April 14, 2026Your Engineering Team's AI Problem Isn't What You Think It Is
Most engineering teams already have access to AI coding tools. That is not the hard part. The real problem is that leadership often treats AI adoption like a software procurement exercise instead of an operating model change. Without clear norms, lightweight governance, and meaningful measurement, AI becomes another line item in the budget instead of a real driver of engineering throughput.
April 1, 2026Engineering Governance Is Not a Committee
Most engineering governance bodies do not govern anything. They approve, notarize, and preserve the appearance of control while avoiding the one decision that actually matters: whether work should stop. Real governance is not a committee ritual or an approval workflow. It is the willingness to kill projects that no longer deserve capital, attention, or engineering time.
March 17, 2026PE Should Require AI Headcount Reduction Targets in the First 100 Days
Private equity firms should stop treating AI efficiency as a vague aspiration and start modeling it as a real operating lever in the first 100 days. That does not mean crude layoffs on day one. It means building a numbers-attached plan for how AI changes engineering team structure, delivery capacity, and cost over the hold period. The firms that sequence this well will outperform. The ones that keep “exploring” will waste precious quarters pretending the math does not exist.
March 17, 2026The Engineer Who Won't Use AI Is the New Engineer Who Won't Write Tests
The engineer who refuses to use AI today looks a lot like the engineer who refused to write tests a decade ago: often smart, often experienced, and increasingly on the wrong side of where the profession is headed. AI tooling is following the same adoption curve as other once-controversial engineering standards, only much faster. In the near future, refusing to use AI will not read as principled. It will read as professionally obsolete.
March 16, 2026Your Engineers Should Be Scared. That's Not a Bug.
AI is not just changing how engineering teams work. It is changing how many engineers companies believe they need. In PE-backed environments especially, the new pressure is structural: leaders are being pushed to define an AI-augmented performance bar, whether they admit it or not. The real problem is not that engineers are scared. It is that too many executives know the bar is moving and still refuse to say where.
March 16, 2026AI Doesn't Make Your Company More Productive. Your Company Makes AI Productive.
AI does not automatically make a company more productive. It magnifies whatever kind of organization it enters. Companies with strong processes, clean data, and leaders willing to redesign work see meaningful gains. Companies looking for a plug-and-play efficiency miracle usually get disappointment dressed up as innovation.
March 16, 2026