Don't use AI to replace people where the thing you're offering is trust: customer service, sales relationships built on accountability, care during a crisis, mentorship, and leadership through uncertainty. In those jobs the person is the service, and automating them makes it worse. Use AI on the work around those people, and only once you understand the process.
Everyone wants to automate. AI is flashy, it's cheap, and it's always there. But some of the most expensive AI mistakes we've seen aren't failed implementations. They're the ones that worked exactly as designed, and still made things worse.
The real issue? AI is a poor substitute for trust.
Why is customer service the riskiest place to automate?
Right now, one of the most popular AI deployments in business is the one least likely to succeed: automated customer service.
Companies are replacing support desks with voice agents and chatbots because the math looks compelling. Fewer people. Faster responses. Coverage around the clock.
But think about why people contact customer service in the first place. Something went wrong. They're frustrated, confused, or out of money. They need someone to hear them, and then fix it.
What they get instead is a bot that mishears their name, loops them through a decision tree, and, if they're lucky, eventually transfers them to a person who can actually solve the problem.
The damage isn't just a bad experience. It's a broken relationship with a customer who was already on the fence. The department businesses are rushing to automate is the one with the most direct line to retention.
Where are people not optional?
AI is the wrong answer anytime the offering is trust, support and authenticity. In those cases, automation degrades the value rather than adding to it. These aren't roles where better technology can fill in for humanity. They're roles where the humanity is the service.
| Work | Why the person is the service | Where AI can still help |
|---|---|---|
| Customer service | The customer needs to be heard by someone who can fix it | Finding account history, drafting replies, routing tickets |
| Sales relationships built on trust | The customer is buying someone's accountability | Research and meeting prep before the conversation |
| Care during critical life challenges | Presence and judgment are what's being offered | Paperwork and scheduling around the care |
| Mentorship and personnel development | Growth depends on a real relationship | Preparing materials and tracking goals |
| Leadership during uncertainty or conflict | People need to hear it from a person who owns it | Organizing facts before the hard conversation |
How do you decide when AI isn't the answer?
AI isn't the answer when the value of the work comes from a person being accountable, present and trusted, or when you don't yet understand the process well enough to describe it. The first test is simple: ask what the customer, client or employee is actually receiving. If it's information, a draft, a search or a routine transaction, AI can often help. If it's being heard, being cared for, or knowing that a specific person stands behind a decision, replacing that person degrades the service, however good the technology is. The second test is clarity. If ownership is ambiguous, the steps live in someone's head, or the output depends on unspoken judgment, automating it doesn't fix it. It locks in the confusion and runs it at scale. Do the harder work first: map what actually happens and where the real value is created. Then use AI to support the people doing the work, not to stand in for them where trust is the point.
Why does clarity have to come before automation?
The bigger issue we see isn't companies automating the wrong things. It's companies automating before they understand what they're actually doing. That's one of the main reasons small business AI projects fail before they start.
The companies that get the most out of AI do the harder work first: mapping what actually happens, identifying where the real value is created, and building from a foundation of clarity. Knowing when not to use AI is itself a sign of skill. On our AI Proficiency Scale, it's part of what defines a Systems Builder: someone who redesigns the work itself, and knows when not to use AI.
That's the part that needs a thinking partner, not a tool.
What can a good AI strategy offer?
This is what our AI strategy and advisory work is built around. Not recommending software. Not deploying agents. Sitting alongside your leadership team and asking the questions that show where AI can help and where it will just get in the way. If you want a structured place to start, our AI readiness assessment lists seven questions to answer before you spend.
Most businesses don't need more AI. They need a clearer picture of their operations first. Once that exists, the right automations become obvious, and the ones that would have caused problems are easy to rule out.
If you're being told that AI should touch every part of your business, you're talking to someone with something to sell. The honest answer is more selective than that, and a lot more valuable. See how we work on our Advise, Educate & Lead page.


