Adam Gonnerman

“The purpose of a system is what it does.” — Stafford Beer A team drowning in slow approvals decides the problem is speed, so they add AI. Leadership announces a new assistant to unlock productivity. A chatbot is rolled out to close knowledge gaps and reduce internal friction. Demos look promising. Early outputs feel impressive. But weeks later, nothing fundamental has changed. Decisions are still unclear. Ownership is still fuzzy. Data is still inconsistent. If anything, the noise level has increased. You can't use AI to repair broken systems. What it does in reality is accelerate their weaknesses. You see, AI multiplies what already exists. Unclear decision rights leads directly to faster confusion. If no one knows who owns decisions, AI generates more options. Further, more stakeholders weigh in, and decision latency increases. Think about it like this: AI increases surface area of disagreement. Poor data result in confidently wrong outputs. If your data are incomplete, incons...

“By failing to prepare, you are preparing to fail.” - Benjamin Franklin In startup and tech culture, we celebrate the last-minute save. But heroics are often symptoms of earlier neglect. Hero Culture thrives on moments like this: it's Friday at 4:47pm, a ticket has been sitting in review for two weeks, the stakeholder is impatient, and the engineer has a flight to catch, so someone pushes the deployment button because it's a small change and it's always fine. By 9pm there's a Slack thread, by 10pm it's a call, the engineer is somewhere over Ohio in airplane mode, and the one person who knows that part of the codebase is stepping outside a birthday dinner every fifteen minutes to check their phone. The fix takes forty minutes, the post-mortem takes two weeks, and the person who stayed up to save the night gets praised on Monday morning while the person who warned against Friday deployments three sprints ago is quietly forgotten. That's the thing about Hero Cult...

“Only when the tide goes out do you discover who’s been swimming naked.” -  Warren Buffett One consistent truth I have observed in business is that leaders tend to reward what is visible, while some of the most essential work within organizations remains unseen. Consider a product launch. Slack channels fill with celebration, leadership offers public praise, and performance metrics are widely shared. Now contrast that with months of refactoring, quiet mentoring, and diligent risk mitigation that largely go unnoticed. Visibility distorts perception. This is visibility bias in action. It appears as: Shipping rewarded more than stabilizing. Roadmaps praised more than operational rigor. Deck-building valued over difficult conversations. Activity mistaken for impact. This connects directly to a post I previously wrote on shipping versus advancing , where I argued that organizations often conflate visible activity with meaningful progress. Shipping, with releases, tickets, a...

Several years back I was in a program role supporting high-visibility digital platform launches. As usual for such a role, I was responsible for driving an on-time release tied to contractual distribution milestones. Meanwhile, a senior engineer was focused on refactoring part of the integration layer to improve long-term maintainability. We were seemingly at cross-purposes, because my goal was to schedule adherence and certification readiness, and the senior engineer was technical sustainability and reducing future defects. These priorities seemed to be in tension, as deeper refactoring risked delaying launch. How could we move forward? Thinking it through, I decided that instead of forcing a trade-off, I would facilitate a working session to break the work into phases. That is, we would identify what refactoring was critical for launch stability versus what could be sequenced into a fast-follow release. Through this process we aligned on a minimal, high impact improvement set that re...

AI adoption in engineering is nearly universal, but the productivity story being told to investors and the one unfolding inside engineering teams are starting to diverge in some uncomfortable ways. In this video, we look at what the research actually shows about AI-generated code, where the hidden costs are accumulating, and why the organizations moving fastest may be building up a problem they won't fully feel until it's expensive to fix.