How to actually use AI in a small business: 10 lessons from the trenches

Follow ZDNET: Add us as a preferred source on Google.

CNBC recently reported on conflicting numbers about AI adoption and productivity among small businesses. Productivity among the S&P 500 rose 5.5% since the introduction of ChatGPT in 2022, according to a Wells Fargo analysis. 

Conversely, the same report says productivity has declined a whopping 12.3% for the Russell 2000, Wells Fargo's name for the bottom 2,000 stocks in its tracking data.

But then, the same article says a QuickBooks survey of 5,000 small businesses shows 68% "integrated AI into their daily operations," but has no details about whether that improved productivity or not.

Your business isn't a statistic, and neither is mine. As I wrote in the companion piece to this article on the 21 ways my business has benefited from AI, it is absolutely possible to benefit from AI use. However, to do so, you must completely disregard the hype, focus on the bottom-line results, and take it slow.

In this article, I'm going to share with you ten key tips that will help you use AI effectively to boost your productivity and your bottom line.

And that, dear reader, brings me to my first tip.

1. Don't spend on AI just because it's the new hotness

Don't invest in AI because everyone says it's the hot new thing. Invest in AI if you see a way it will bottom-line benefit your company. You wouldn't spend on a piece of equipment just because it's cool. You'd buy it because it opens a new business opportunity, helps you improve productivity, or helps you reduce costs.

AI should be treated exactly the same way. If you don't know how AI will help, don't spend a ton on AI. In fact, don't spend anything until you do know how it will help. 

2. Experiment using the free AI chatbots

I recently conducted objective performance testing of the eight most popular free AI chatbots and was blown away by the results. The best ones (ChatGPT, Copilot, Grok, and Gemini) have improved tremendously in the past two years. They're not perfect, but they definitely get the job done, especially when you're just tinkering and trying them out.

Until you find at least one direct benefit of using AI in your business, keep using the free AI tools.

3. Spend monthly, not yearly

Let's do one more money tip, and then we'll move on to the more operational tips. AI costs are direct functions of how many resources the AI companies have to provide for a given response. It's entirely fair for AI companies to charge for resources, especially since the industry is spending hundreds of billions (with a "B") of dollars building the data centers that make this possible.

That said, until you're completely sure you want to integrate a given AI service into your business long-term, don't sign up for a yearly commitment. Some AI services do require that year commitment. If you need those, you'll have to think it through. But most let you buy monthly, albeit for a few more bucks than a yearly commitment.

Also: I got 4 years of product development done in 4 days for $200, and I'm still stunned

When I spent $200 on OpenAI's ChatGPT Pro tier, it was a lot more than my usual $20/month spend. I used it for a month. During that month, I was able to crank out four full products.

Then I went back down to my normal tier of service because, at this point, I don't need to generate more products. Instead, I need to integrate and market the ones I do have. I don't see myself needing to plunk down another couple of Benjamins for at least six months, if not longer.

So, instead of spending $2,400 for a year, I spent $200 for a month. Win.

4. Define what constitutes a win

For me, it's tangible time savings. I operate a two-person professional firm, which means our income is directly related to the amount of work product my wife and I can produce.

For years, we ran a typical startup with employees and commercial property, where growth was paramount, but we missed doing projects and didn't enjoy all the overhead involved in managing employees. We retired that operation a few years ago and now run a content-based business out of our house.

Practically, that means the more we can do in a given amount of work time, the better. For us, a win is when we can invest a little cash and get it back in time.

For example, every month I create a specialized image that's an involvement device for my wife's e-commerce operation. It used to take me about two hours or so using Photoshop. But when I added in the $10/month subscription to Midjourney (an image-generating AI), I was able to get the job done in about 10 minutes.

Paying ten bucks to reduce a regular project from two hours to ten minutes is a win. It's not about using AI; it's about recovering 110 minutes every month on this one regular task. That's how you look at AI. What are your tangible wins?

5. Use the AI for quick analytics reports

You can feed raw data into the AI for quick analysis. I did this for a sentiment analysis about products a few years back, and that helped inform my product strategy. However, today, as I'm writing this, I decided to explore what AI can do with a standard QuickBooks profit and loss report.

I used QuickBooks' reporting tools to generate a report showing the yearly P&L for the last five years, exported it as a PDF, and imported it into the free tier of ChatGPT. I uploaded the PDF and asked, "What can you tell me about this?"

It gave me an overview, an income breakdown, an expenses breakdown, a profitability trend, and some key takeaways. It also asked, "Would you like me to create a visual analysis (charts/graphs) showing income, expenses, and net income trends by year?" I answered yes and got a nice series of charts.

Then I asked it, "What unexpected or hidden insights can you find in the data?" I'm not going to share the results, other than it noticed we spend a lot on digital infrastructure (which makes sense given our business), but it proved to be fascinating.

My tip here is to take data that you might not otherwise have the time or resources to analyze, give it to the AI, and see what it reveals. There is one caveat: The AIs are not supposed to share that data, but you never know. I think the risk is the same as running your email through Google or keeping your Office documents in OneDrive. There is risk here, but there's always risk in allowing cloud services access to your proprietary data. It's the world we live in.

6. Pair your expertise with the AI, don't replace it

Don't look to use the AI for headcount reduction, especially if you're a small business. The AIs can do a lot, but they're also classic management challenges. Think of them as very bright but annoyingly troublesome helpers. To succeed with them, they require constant guidance and course correction.

There's a business school concept generally known as the "degree of autonomy," which describes how long an employee or staff member can operate without direct supervision or check-in. You know the case: Some employees you need to check on every hour or so, while others can handle an assignment and just come back in a week when it's done. Chatbots should have a very low degree of autonomy because, from one prompt to the next, they'll go off the rails.

That's why pairing with expertise is powerful. If you manage the tools as you use them, you can gain astonishing productivity benefits. But if you just assume they'll work or use them to replace human workers, all that productivity is likely to be lost cleaning up the messes they will inevitably produce.

7. Avoid AI coding unless you're a coder or have a coder on your team

Look, AI coding can be astonishing, game-changing, and have an ROI almost beyond comprehension. However the AI coding tools can also be incredibly stupid, thickheaded, stubborn, and take exactly the wrong approach to any problem.

Vibe coding has been a big buzzword this past year, inspiring non-programmers to think they can just talk their favorite AIs into creating multimillion-dollar apps. I mean, there's always a chance. But it's unlikely, especially if you run into a snag or need to fix something the AI is doing wrong.

Also: 10 professional developers on vibe coding's true promise and peril

I've personally proven that AI coding can be an incredible boon to programmer productivity. But I'm a programmer. If you have a coder on staff or are one yourself, don't hesitate to invest in AI coding tools. But if you do not have that experience or training, you're just setting yourself up for a world of hurt.

8. Use AI to brainstorm

The AIs can be great brainstorming partners. I've used the AI to brainstorm product names, YouTube headlines, character names, character backstories, social media tags, and more.

I've also used AI to set up straw man proposals for new business projects. I've even done it for shop projects. For example, I fed ChatGPT Plus a photo of my workshop and then asked, "I'm thinking of adding a winch to this beam. Straw man it for me." I got back a detailed analysis, and then, when I asked for images, it even drew some diagrams with recommendations.

Just be careful. The AIs are fairly bad at drawing engineering diagrams from text-based prompts. They can make things that are pretty or seem real, but they're not good at diagramming concepts. Sometimes, though, you can critique drafts and they do get better. Other times, that process results in completely ridiculous diagrams like this one.

9. Start small, scale gradually

This piece of advice isn't new for any business endeavor. We've talked about it for setting up cloud services, building out IT infrastructure, and more.

Adding AI is a process. You need to get to know it -- its strengths and weaknesses -- before you do a lot. See what works for you and what doesn't. Keep an eye out for those things (like that monthly image I have to produce) that are tangible wins, or anything where you're spending a lot of time and the AI might be able to step in.

This applies to your team as well. Let them experiment, but be sure they're testing before they're deploying. One piece at a time, get to know its characteristics, and build on that.

As the companion article to this piece shows, I can point to 21 different ways my little company has benefited from AI over the past few years. But I didn't do them all at once. Each was attempted and tested one at a time.

10. Value time as your key metric

When we make business investments, we usually measure ROI (return on investment) as our key metric. We put in X dollars and got back Y dollars. But, at least initially, you're not going to see easily attributable dollar returns to your AI investments.

Either that, or you'll see ludicrously high perceived returns on tiny $20/month investments. Neither can really be properly measured for future evaluation.

I recommend, at least for your first year or so with AI, that you focus on time return. How much time did this AI project save? How much time did this AI project free up? What did you get done in a much shorter time than you otherwise would have?

If you follow my guidelines and use the free plans to start, and you see time savings you can attribute to the AI, then you can begin to guesstimate how much value AI provides to your business.

Personally, I've seen a few tangible, repeated time savings that more than justify the investment, one absolutely bonkers ROI (the $200 programming sprint), and a ton of tiny five-minute here, ten-minute there chunks of time saved.

What about you?

I didn't go into many specific examples about how my small business uses AI because there's a companion article that details 21 specific examples. I recommend you read that.

Beyond that, keep in mind that each business is different. You might have a one-person business or a hundred-person business. Each will benefit from AI differently, but there's no doubt that there are some upsides to adding AI to your set of available tools and resources. Just be smart about how you use it. Don't fall into the trap of thinking you can substitute some AI service for smart and capable humans.

What about you? Have you started using AI in your small business or personal projects? How do you decide when a tool is worth the investment versus just hype? Have you found any specific AI tools that truly save you time or improve results? Or are you still experimenting to find the right fit? Let us know in the comments below.

Want more stories about AI? Check out AI Leaderboard, our weekly newsletter.

You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.

Artificial Intelligence