In Part 1, we looked at how AI mania has scrambled decision-making inside a lot of companies: projects that quietly fail, a workplace where doubting AI became a career risk, demos that override good judgment, and executives too boxed in to say what they actually think. Underneath it all, the confidence you hear about AI is often manufactured, and the people who know better have learned to stay quiet.
Here’s the good news for a small business: almost none of that machinery is running inside your walls. This part is about the advantage that gives you, and how to use it.
When the Label Matters More Than the Work
Put all of this together and you get organizations where the AI label has become a gate every project has to pass through, whether or not AI has anything to do with the job. Sensible work gets dressed up as an AI initiative to win approval, and useful work that can’t be dressed up gets quietly starved.
Consultants describe watching routine projects get relabeled this way. In one case, a company needed to move off an aging system before a costly license renewal, an ordinary migration. The vendor bolted an AI step onto the front of it, and when that step didn’t work, they finished the job by hand and reported it upward as an AI-driven success, because a sliver of the work had briefly passed through a model.
This isn’t rare, and it isn’t only internal spin. In 2025 the FTC brought a series of “AI washing” cases against companies that had marketed products as AI or machine learning when the real work was being done by hand. When a label is worth more than the truth, the label is what you get.
The pressure runs the other way, too. Work that can’t wear an AI badge gets held up, reworked until it sounds “AI enough,” or denied outright. On the hiring side it has been made explicit: in 2025 Shopify’s CEO told staff that before requesting more people or budget, teams had to first prove they couldn’t get the job done with AI. In one of the technology communities that I’m a member of, the person who runs it cited surveys that showed that hiring managers prefer to hire developers with less experience coding who also had AI skills than those with more experience coding who had no AI skills. The effect is that the label, not the goal, starts deciding what gets built and who gets hired.
None of that machinery is running inside your business. You can still do the thing these large organizations have lost the ability to do, which is judge a project by whether it moves the needle for you. When you’re paying someone to build or rebuild something for your business, judge it by whether it solves the problem you have and holds up over time, not by whether it can be called AI.
A tool that quietly does its job every day is worth more than one that wears a badge to impress a boardroom.
Where AI Helps
After all of that, it’s only fair to show where AI does deliver, because it does. The tools are genuinely useful on the right kind of task. The trouble only starts when a narrow, capable assistant gets mistaken for a company-wide strategy.
The wins tend to share a shape. AI does its best work on a single, well-defined job, with a person who can judge the result. A few that hold up for a small business:
- Turning a blank page into a first draft. For a routine email, a service description, or a rough how-to document, AI is good at getting words on the page that you then shape into your own voice. This is also how you finally write the internal documentation whose absence quietly sinks so many company chatbots.
- Making sense of material you already have. Point it at a long contract, a pile of meeting notes, or a recorded call, and it can summarize it or pull out what matters. It’s working from something real you handed it, which is when these tools are at their steadiest.
- Handling the fiddly parts of everyday tools. Ask it for the spreadsheet formula you can’t remember, to explain what a sheet is doing, or to tidy a messy export. You can see straight away whether the result is right.
- Speeding up the developers you work with. A capable developer can lean on AI to move faster through routine code, because they can read the output, catch what’s wrong, and stand behind what ships. Their skill is what makes the tool safe to use.
- Triaging customer messages, with a person behind it. AI can sort incoming questions and draft first replies, as long as a real person is one step away for anything it can’t handle on its own.
None of these tries to reinvent the business. Each one takes a single task off your plate and leaves you in charge of the result. That’s the same thing the MIT researchers found in the small share of AI projects that paid off. Those projects picked one real problem and solved it well, rather than rebuilding the whole company around the technology.
Pointed at one job you can check, AI is a useful assistant. Turned into a strategy, it becomes the mania everyone else is caught in.
Keep a Clear Head About AI
None of this is a reason to swear off AI, and it’s not a reason to worship it either. The useful stance is the one these big organizations have lost, which is staying in a position to judge each tool by what it does for your business. The mania makes that hard for them. It doesn’t have to be hard for you.
A handful of habits keep you there.
- Start with the problem, not the tool. Don’t go hunting for places to add AI. Begin with a real bottleneck, a slow task, a cost that won’t come down, a lead you keep losing, and ask what fixes it. Sometimes the answer is AI, and often it’s something simpler.
- Judge it on a bad day. Before you commit, run the tool on your own real work and watch what it does when it’s wrong, on your data, before its in front of the customers you can’t afford to lose.
- Look under the label. Plenty of tools wear an “AI-powered” badge over ordinary software, or over people doing the work by hand. Make sure you’re paying for capability, not branding.
- Keep a person reachable. If you put a chatbot or an automated line in front of customers, make sure a real human is one step away and nothing quietly drops into a void. A tool that loses you customers while looking fine on a dashboard is worse than no tool at all.
- Own what you build. If you do build on AI tools, make sure the result is something you understand and can hand to any developer. A prototype that ships fast is only an asset if you’re not locked out of it later.
A lot of small businesses get caught with that last bullet. Something gets built in a hurry, with AI tools or without, and a year later it’s slow, fragile, and impossible to change, with no one quite sure how it works. Rebuilding it properly, in clean code you own outright, is often the difference between a tool that grows with your business and one you spend your time babysitting.
You don’t need to match anyone else’s pace, and you don’t need an opinion on every headline. You need your business to work, and you’re free to use or ignore AI based on what actually helps your business.
The mania will pass, the way it did for blockchain, the metaverse, and NFTs. The habit of asking whether a tool earns its place is worth keeping long after it does.