a practical guide to using ChatGPT, Gemini, Claude, and more

You’ve probably heard the saying, AI will not replace you—but a human using AI will. After all, it’s the sentiment expressed by CEOs, the likes of Nvidia’s Jensen Huang, Indeed’s Chris Hyams, and Netflix’s Ted Sarandos.

However, getting ahead of the game and implementing AI into your work routine is often easier said than done—especially for those who don’t even know where to start. In fact, getting up and running is one of the biggest challenges in today’s AI era, according to Jules White, a computer science professor at Vanderbilt University.

“The main problem is that most people don’t know where to start when confronted with a blank generative AI canvas,” White tells Fortune. “The limit isn’t the technology, the limit is our ability to creatively use it in our work and daily lives.”

There’s no question: AI can make the lives of workers easier and more productive—whether it’s to summarize long documents, generate slideshows, analyze speaker notes, or something much more advanced. Here’s what you need to know to make it happen—and effectively prompt generative AI chatbots like ChatGPT, Gemini, Claude, and Copilot in the workplace:

Follow the golden rule of prompting

Growing up, you probably were told that the golden rule was to treat others the way you want to be treated. And in a sense, AI operates in the same way: treat the technology just like you’d ask someone for something—making sure it has enough context to know what’s going on.

Anthropic sums it down to the “golden rule of clear prompting:”

“Show your prompt to a colleague, ideally someone who has minimal context on the task, and ask them to follow the instructions. If they’re confused, Claude will likely be too.”

In practicality, there are four main areas of focus when writing a prompt, according to Google’s Prompting Guide 101: Persona, Task, Context, and Format. Here’s a simple example it cites:

“I’m a PR manager. (Persona) I need to create a press release with a catchy title. (Task) Include quotes from (Format) @[Document name] (Context).”

Not every prompt must include the four areas, but providing effective context helps move the needle when it comes to the response, according to Maggie Vo, head of user education at Anthropic—the company behind Claude.

“Instead of ‘Write a summary of quantum computing,’ try ‘I’m preparing for a job interview at a quantum computing startup. I have a physics background but no quantum experience. Help me speak intelligently about key concepts,’” she suggests. “That context completely transforms the output.”

Following the golden rule also might mean you are tempted to be polite to AI by saying please and thank you—something OpenAI CEO Sam Altman has said is costing his company “tens of millions of dollars.” But in actuality, there’s little indication social niceties significantly improve outputs.

The persona pattern—meaning assigning the AI a specific hat or writing style—in particular is a great gateway into learning how different AI is, White says; in fact, it’s the very first thing he teaches in his class, “Prompt Engineering for ChatGPT,” a Coursera class that’s been started by more than 450,000 individuals.

“The persona goes way beyond things like ‘act as a professor of computer science,’” he tells Fortune. “You can ask Generative AI to ‘act as a computer that has been the victim of a cyber attack’ and it can simulate how that computer might work.”

Above all, Google notes that including a verb or command as part of your task is the “most important component of a prompt.” Unlike a traditional Google search, which might assume you want to purchase “cat food” if that’s all you type, typing the phrase into AI might leave you will more questions that answers:

Treat AI like a fostering a relationship

Even with a seemingly great prompt, chatbots may not give you the response you will be looking for. While this might feel especially frustrating after utilizing a deeper reasoning model that took several minutes to process, it shouldn’t necessarily be a reason to slam your laptop shut.

Rather, use it as a learning experience. Though not all AI have persistent memory—and will learn from your interactions—your prompting expertise will only improve with practice.

As Google states, “Fine-tune your prompts if the results don’t meet your expectations or if you believe there’s room for improvement.” But this doesn’t mean you necessarily have to start a new session and copy your old question over again. Instead, “use follow-up prompts and an iterative process of review and refinement to yield better results.”

In practice, this also means you should point out errors or lapses in judgment from the responses because while hallucinations are not as prevalent as they once were, perfect AI accuracy is not guaranteed.

Say you ask for the AI to give you the estimated population of 50 cities in the U.S., and the output mistakes Portland, Maine, with Portland, Oregon—or it skips a city altogether. Instead of just repeating the prompt again until it gets it right, you could simply point out that it skipped over two entries—and the AI will realize its mistake.

In a sense, interacting with AI should not feel like a monotonous, copy and paste-type relationship, White says. Instead, use it like a partner.

That can include you asking the AI questions. For example, “I need help writing a marketing email, what details would be helpful for me to supply?”

And as Ethan Mollick, a professor at the University of Pennsylvania’s Wharton School—who is known for his prolific AI research and analysis on LinkedIn—points out, don’t be afraid to push the AI to its limits.

“Ask for 50 ideas instead of 10, or thirty options to improve a sentence,” Mollick suggested on his Substack. “Then push the AI to expand on the things you like.”

Mollick did not respond to Fortune’s request for comment.

Use more than just the chat function

AI innovation has surpassed just text-based interaction. You can upload spreadsheets, slideshows, and markup files for analysis. Say, for example, you have a spreadsheet full of recent sales data—instead of spending time trying to create pivot tables—AI can be asked directly, “Which sales region had the highest revenue growth over the last two quarters?”

Certain chatbot models, like ChatGPT and Midjourney, also have the ability to create graphs, slideshows, and pictures—or analyze ones you upload.

“My wife can snap a picture of the random ingredients left in our fridge and get 10 recipes that she could make with them,” White says. “She can turn around and have it plan meals for a week.”

And if you aren’t a fan of typing, many chatbots have voice capabilities that could be a gamechanger in the business world by allowing you to not only have a seemingly natural conversation—but also share your screen or view your camera and react in real time.

“I’ve used it to identify plants on hikes, solve a problem on my screen, and get cooking tips while my hands were covered in flour,” Mollick wrote. “This multimodal capability is genuinely futuristic, yet most people just use voice mode like Siri. You’re missing the best part.”

Don’t overthink it

Practice makes perfect when it comes to AI technology. Becoming an expert prompter or “AI whisper” is no longer a skill with a large barrier to entry. In fact, in today’s era, the most effective tactics revolve around using natural language, just like you’d use when asking a co-worker for help.

And worse-case scenario, the best way to learn how to prompt is asking the AI itself.

“AI can literally teach you how to communicate with it better,” Vo says. “It’s surprisingly effective and saves you from memorizing prompt templates.”

Lastly, make sure to ask your employer if there’s any guidance around using AI at work. If not, ask, why not?

This story was originally featured on Fortune.com