Over-Automating
There’s a version of AI adoption that quietly automates you out of your own business. You start removing yourself from processes. You hand more and more off to systems. And before you know it, the human touch that made your business trustworthy in the first place has been engineered away. Your clients aren’t sure who to thank. Your team isn’t sure who makes the call. And you’ve optimized yourself into irrelevance.
Here’s the thing about AI that doesn’t get said enough: AI is not accountable. Humans wielding AI can be accountable. That distinction matters enormously in service businesses. People want to be able to thank someone when things go right. They want to be able to call someone when things go wrong. They want a real human being on the other end of the relationship, someone who will answer for what happened. No amount of automation changes that desire. If you automate your way out of being that person, you’ve eliminated the thing that actually made you worth trusting.
The better frame is: what do I want AI to make easier so that I can be more present in the ways that matter? In a security company, for example, AI making it easier for clients to know their staff is performing, making it easier for managers to see who’s on-site and on-time, making it easier to track equipment and uniforms: all of that is good. It frees up mental bandwidth for the human work of building client relationships, resolving real problems, and making judgment calls that no system should ever be making. The technology serves the human layer. It does not replace it.
There’s also a positioning issue to think about carefully. If you go around telling people your service runs on AI, you’re planting a seed of doubt you didn’t need to plant. The better story is: we use the best tools available to make sure our team shows up and performs at the highest level. AI is infrastructure, not identity. The client cares about outcomes. The client cares about trust. Let the system be invisible and let the relationship be what they remember.
One of the best illustrations of this came from Ryland, the founder of SwitchBooks (an agentic replacement for QuickBooks), on a panel at SXSW in March 2026. He built his agent system and, over the course of a week, his codebase ballooned to 100,000 lines of code. The agent was trying to account for everything a user could possibly say, which overloaded the context window and made the whole system fragile. He stepped back, realized the architecture needed to be rebuilt around simplicity, and rewrote it. The result: about 200 lines of code that performed, in his words, “literally a 1000 times better.” His insight was sharp: “No one wants to be overloaded with information. What people want is a finite set of rules. And then they want to use their reasoning on their own.” That principle applies equally to AI systems and to the humans running them. Over-automating is not just about doing too much. It is about giving the system too much. The constraint that produces great work is clarity, not volume.
Key Takeaway
Automate to free yourself up for the human work that builds trust, not to replace the human layer that makes your business worth trusting in the first place.