Guest commentary: How to master AI without losing the human touch
By Jim Sterne
The hardest part of adopting generative AI is identifying the use cases. That’s why most people are still stuck on “help me write this email.”
Anthony Hamilton is well past that. He filled out a Better Business Bureau survey about AI use and his answers were interesting enough that I asked for a call.
Hamilton co-founded THE Water Heater Company in Camarillo in 2019. Today, the company has 25 employees, including 13 field technicians, and is looking to hire seven more to serve Southern California from Santa Barbara County through Orange County — for now.
The company’s approach to AI is deliberate: automate internally and keep humans in front of the customer. They took a shot at having AI answer the phones and quickly discovered the true value of human connection. Customers did not like talking to AI.
“We tested it, and it did not provide the level of quality that we expect for our client experience,” says Hamilton, indicating a willingness to experiment and learn quickly.
The AI chatbots are gone, replaced by an automated email explaining who they are and promising a human follow-up. “People can be hard on a human, but they’re really hard on the AI. The expectations are through the roof.”
That matches my own experience. Apple’s AI assistant took my complete-sentence explanation of what I needed, summarized it, and handed me to a human in five seconds. I didn’t have to repeat myself, and didn’t have to trust a machine’s product recommendation. As AI handles more of the interaction, human attention gets rarer and therefore more valuable.
So where does AI shine?
Their new website boasts almost 1,600 pages, with a dedicated page for nearly every city served. Why the specific pages? Local municipal water quality varies from city to city, which informs what kind of heaters, filters and softeners are appropriate in each location.
Over 3,200 customer reviews offer superior social proof. Pop-ups showing actual installations completed in real time provide more social proof and create an accelerated sense of urgency.
The company’s good-better-best approach to upfront quotes explains why revenue is expected to double next year. Hamilton spent several months collecting the data, but his web design company spent just days putting it together because he had so assiduously defined what he wanted.
Internally, AI writes memos, SOPs, product brochures and catalogs. Job applicants complete a 100-question assessment, and AI helps evaluate their responses against a rubric informed by research on high-performing employees.
Screenshots of customer reviews identifying specific technicians get assembled into branded PDFs by an office coordinator using a design system Hamilton built in Claude.
Documenting how work gets done is the bottleneck in almost every company because the people who know the process know, instinctively and intuitively, how to get the job done. It’s like muscle memory, so getting them to explain it in detail can be painful.
Hamilton uses dictation software and narrates each task out loud while he’s doing it. He then feeds the transcript to an AI skill that enforces a consistent SOP structure.
The hardest part is keeping up with the speed of change and the number of tools. “This is a recipe for disaster: learning constantly, shipping nothing. People need to sit down, without AI, and figure out what their priorities are. Then take that and plug it into AI.”
Hamilton wants an AI technical support agent trained on his product manuals and his own support call transcripts, capable of taking unlimited simultaneous calls from technicians in the field, because his office team won’t have the capacity to support a doubled field team. Eventually, he also wants an orchestrator that delegates tasks to other agents.
The goal isn’t headcount reduction. “We can enhance ourselves with AI but not replace ourselves,” Hamilton said.
The second hardest part is change management. Helping team members become comfortable with new tools requires patience, context and a light touch. People naturally resist technology when it feels imposed on them or threatens their sense of competence.
“I’m optimistic that our early adopters of AI will help demonstrate to the rest of the team how these tools can make their work easier and help them become more productive,” he said.
For Hamilton, the lesson is simple: AI should remove friction, not humanity. The goal is to build a company where every person can do more meaningful, higher-quality work. The companies that benefit most from AI will be the ones that understand their priorities, document how work gets done and use technology to make good people better.
• Jim Sterne is a co-founder of Coastal Intelligence in Santa Barbara, an AI think tank and consultancy organization.
