Skipping AI Feels Low-Risk. So Did Skipping Email in 1995.

In 2024, the American Bar Association reminded its members that a lawyer’s duty of competence is measured against the tools of the day.

A lawyer in 1990 could skip email and still serve clients well; a lawyer today cannot. As tools improve, the floor for “competence” rises with them.

What interests me is that the logic extends beyond the law into small business.

Lawyers aren’t the only professionals held to a standard that moves. Registered investment advisers carry fiduciary duties, real estate agents owe them under most state laws, and accountants are measured against professional standards of care. So are the trades: a master electrician is judged against today’s code, not the code in force when she apprenticed. Whenever a duty of competence meets evolving tools, the same dynamic applies: The floor rises whether you’re watching it or not.

The Small-Business Constraint

Big companies staff a whole crew for changes like this: an innovation team to scout ahead, an IT department to work the rigging, and a compliance officer to steady the wheel. A solo practitioner or small business has none of those luxuries. You have a business to run and clients to serve.

Questions about whether to adopt AI and which tools to try sink to the bottom of a very long list of things to do, where they feel low-risk because right now, nothing’s on fire.

What causes the demotion is understandable: the uncertainty about value, effort, and downside. But I would argue that the bottom is the wrong place for those questions, and low-risk is the wrong category.

‘I Don’t Want AI in My Business’

Almost no one wants AI in their business until they see, firsthand, how useful it can be. I beg you to find the hour it takes to try one of the models.

Generative and agentic AI is moving fast. Compute is scaling in months, not years; model quality keeps climbing; and meaningful updates now arrive every few weeks. For some narrow tasks, what was clumsy last quarter is reliable this one.

For small businesses that don’t use these tools, the risk is a widening skills gap and what I see as an existential threat. As the standard of competent practice rises (measured by reality, not hype), your relative capability declines. By the time you, your client, or a regulator notices, the gap is difficult and expensive to close.

Start with an Experiment

So, close the gap deliberately.

If you haven’t touched AI recently (or at all), start with a low-stakes trial. Try Claude, ChatGPT, Perplexity, Gemini, or Microsoft Copilot; I have my favorites, and you’ll quickly find yours. Spend the $20 a month for a paid tier upfront and track the time you save. If a tool reliably saves an hour or two a week or makes what you do more effective, it has earned its keep.

Subscription AI gives you access to more powerful, more capable models, which serious work requires. Yes, you’re adding another tool you need to budget for, but what you will get for that money is business-changing. I’ve long thought that one of the reasons business owners don’t adopt AI is because their experience is limited to the free models, which have more limited capabilities and just aren’t as good. “I tried AI once, and it sucked” is a fair statement when all you’ve had access to is free ChatGPT.

When I first started using AI (Jarvis, at the time), I ran an experiment. I asked Jarvis to write a sales email based on the key points I wanted to make. It drafted, I edited, and I was done. Twenty minutes, end to end, for work that would normally take an hour or more. I used that email as a template for months.

That remains the right starting point. Pick something you were already going to do (an email, a meeting summary, a proposal, a follow-up after a client consultation) and hand it to the AI tool after you’ve written it. Read what comes back, keep what you like, remove what you don’t. The tool is how you start without the dread of a blank screen or acts as a second pair of eyes; you remain the arbiter of quality.

Of course, don’t put client-identifying information into these unsecured tools. Validate AI output (it can be confidently wrong), and disclose use when it affects deliverables. Don’t use AI with sensitive personal data, for tasks requiring access to private databases, or for citations or information that requires up-to-the-minute accuracy.

Starting small like this teaches you, concretely, what the tools are good at, which is the only way to find out. Once you have that understanding, the next use case suggests itself: perhaps a first draft of a proposal instead of just a polish pass, or a summary of a long document so you can decide whether to read it closely. That’s how an occasional win becomes a repeatable workflow.

This week, pick one routine task and spend 20 minutes on it with an AI tool after your first pass: Turn notes from a customer meeting into a clear follow-up email with next steps, as an example. Keep only what improves the work, and write down the procedure so you can repeat it.

You aren’t discarding hard-won processes, and you aren’t throwing yourself on the tender mercies of a tool you don’t yet trust or find value in. You’re spending 20 minutes on work you were going to do anyway.

Don’t let this sink to the bottom of the list with everything else that isn’t on fire. Put the experiment on this week’s schedule, and find out where the floor is before a client or a regulator notices first.

Want help sorting this out for your small business? First consultation’s free.