Artificial intelligence (AI) applications can do amazing things, from improving cancer screenings to combating climate change. Little ole you and me, we probably have more modest ambitions, but we still want to get the most from this technology. How do we do that without having to become an AI coding expert?

Fortunately, it doesn’t take much to wring value from AI. We just need to broaden our understanding of its capabilities, recognize its limitations, and follow a few best practices.

AI abilities are expanding

AI is great for writing ad copy, a speech, a sales pitch, or an essay. You can also use it to create a business plan, compose, and send an e-mail message or a social media post, generate website code, or create images, videos, and podcasts. Ask it for suggestions to improve your resume, or to convert rough notes into an outline.

But it pays to get creative. Some less obvious uses may inspire you to dream up your own ideas:

• Translate genealogy records: I used Google Gemini to translate some Latinate text from 13th-century church records about my family into modern, understandable language.
• Fix mistakes: I uploaded an image of a birthday cake with a misspelling on the icing to Gemini, and instructed it to make corrections. Gemini returned an updated image with a seamlessly corrected spelling so we could share it online without having to redecorate the cake.
• Build a business case: I asked ChatGPT to give me talking points about why our company should install electric vehicle chargers, and it gave me 10 clear points supporting that point of view. Then I asked for 10 points about why we shouldn’t install chargers, and I received 10 points supporting that point of view. The CFO never knew what hit her.
• Help with taxes: Scan and upload documents and then extract data like expenses (but remove all personally identifiable information first).
• Get a green thumbs-up: One redditor uploads images of their plants, then asks AI whether they are healthy, and which fertilizer to use.

You can find many other ideas by, well, asking your favorite AI app or using an old-school Internet search. I found many great ideas in a ChatGPT reddit group; you can find AI prompt examples in Google’s library, too.

AI best practices and tips

Match the AI to the task

You wouldn’t edit a photo in Microsoft Word. Likewise, some AI chatbots are intended for generating text, others for creating images or video, and still others for coding. What’s more, one chatbot may be better than another at a particular task. Pick an AI chatbot to start with and be ready to play the field. Another chatbot may produce different results while providing a way to check its competitor’s work.

When I asked Gemini to find restaurants with private meeting rooms in my area, its results were just plain wrong; its recommendation to “check their websites or contact them directly” didn’t help either. Grok gave me several options, but the top one has been closed for a couple of years. ChatGPT gave me great recommendations, many with information about room size or maximum number of guests, plus a map and links to directions, websites, and phone numbers.

You may need to use an AI app that is designed for a narrower function. For example, to record and transcribe phone calls or meetings, try Otter.ai, or the Google app, Braindump. Many directories of AI tools can be found online to help you find the right tool for the job.

Ask AI-friendly queries

Unlike your friends and coworkers, a chatbot can’t guess what you mean when you ask it a vague question. Specifying context and parameters, and using an AI-friendly format, will generate more specific, relevant, and accurate results.

• Context: When I asked ChatGPT where the Oakland A’s baseball team plays, it told me it was at the Oakland-Alameda County Coliseum. That’s no longer true. But when I asked where the A’s play in Sacramento, California, it was able to identify the current, correct ballpark name.
• Parameters: If you are creating a work report, specify the audience, as in, “Create a report for C-level executives.” For an essay, request an “informal,” “conversational,” or “academic” style. To make sense of scientific or medical information, add “explain in layperson’s terms.” You can make the output sound more like you and less like a generic robot by uploading documents you’ve written and asking the chatbot to match your writing style.
• Format: If you have a complicated query, try using a prompt structure like Role, Task, Format to provide clarity. For example: “Act as a finance manager to create a plan to refinance our company debt and format it as a presentation.”
• Use your manners: Microsoft says, “using polite language sets a tone for the response,” and that LLMs are trained on human conversations. Politeness may produce better results, too. On the other hand, OpenAI CEO Sam Altman says that users who say “please” and “thank you” to ChatGPT cost the company millions of dollars in electricity costs. Perhaps you could split the difference by using “pls” and “ty.”

If you don’t like the results, revise your query. Most chatbots will continue to follow the thread you were working on, so you don’t have to start over. For example, an image of people I requested from ChatGPT had bad lighting; I asked it to revise the image to have more detail, and it did.

Protect yourself from AI risks

Respect ethical boundaries: There are many thorny legal issues concerning the use of AI, including the alleged use of copyrighted material to train AI models, the impact of AI bias, and proposed liability rules. Businesses need to be aware of the risks and have compliance policies in place to avoid them.

At least you don’t have to worry about AI leading you down the wrong path: None of the AI chatbots I tried would tell me how to rob a bank, though Grok volunteered to help me write a novel about it.

Create guidelines: Do you want your employees to use ChatGPT to write their communications with customers? Maybe; maybe not, but you’ll need to make decisions like that and then implement training and governance to make sure AI is used properly.

Check their work: Some businesses, and even government agencies, have been caught using fabricated AI-generated content. While AI apps work by accessing massive amounts of data housed in large language models (LLMs), that data might be outdated, incorrect, or even fictitious, so visit the sources they cite and verify the information, especially for important projects.

Private and secure: If you upload anything to a general-purpose AI chatbot, it will likely be stored in the cloud, become owned and managed by the chatbot operator, and may be used to train their models. So, it must be said: Don’t upload personally identifiable information, and certainly don’t upload confidential work-related data. Gemini goes so far as to caution against uploading raw banking transactions—ones that contain routing and account information.

Even if what you do with AI isn’t sensitive, you should still take steps to secure your interactions by adjusting the security settings within your chatbots.

Some enterprise AI apps, including SAP’s Joule copilot app, access only corporate data that is secured on private cloud servers. Only SAP customers can access Joule, and the only private data they can access is their own. Joule doesn’t allow file uploads, and customer data isn’t used to train the models it runs on.

Run AI locally: Outside the workplace (or if your organization lets you), you can download a free app like Ollama or gpt4all and run AI models on your personal computer. Your data will remain local, though your queries will run more slowly than they would in the cloud, where they benefit from AI companies’ huge, nuclear-powered data centers.

AI is learning—how to keep pace

Still want to learn more about how to get the most out of AI? Take online AI courses. Many are free, though if you want to crow about it on your resume or your LinkedIn profile, you may need to pay for a digital certificate.