These days, anyone with an internet connection can tap into the power of generative artificial intelligence (Gen AI) to answer complex questions and generate unique copy, among other things. The trick is learning how to channel that huge wealth of information for your very specific needs, by writing prompts that elicit relevant, appropriate responses.
Such is the role of prompt engineer (or AI whisperer, if you must). But you don’t have to be a pro to craft effective Gen AI prompts. The tips below can help even novice users get better results from Gen AI.
1. Be precise, and give lots of context
A Yale Professor (and no, not Einstein) once said, “If I had only one hour to solve a problem, I would spend up to two-thirds of that hour in attempting to define what the problem is”. This was before computers, but he might as well have been talking about AI.
As with any business writing – or any question, really – clarity is paramount. A vague, unfocused prompt may elicit an equally disappointing response that doesn’t quite hit the mark. For example, if you asked your Gen AI tool to “Write a summary of our annual report,” you might receive a response like, “Our annual report for this year covers our financial performance, achievements and future plans. We have made significant progress and look forward to continued growth.” But that doesn’t say anything specific about your annual report or the results you want to highlight.
Instead, begin with a clear, precise prompt – including lots of context and detail about what kind of output you want to receive – to get a more accurate, relevant answer that better aligns with your needs and expectations.
2. Know what you’re working with
We’re all familiar with OpenAI’s ubiquitous ChatGPT, but there are many different AI tools out there, like QuillBot for academic use, Copy.ai for marketing communications and Codex (also by OpenAI) for coding. Like any tool, each one has specific strengths and weaknesses and will give results designed for specific uses, so you’ll get the best results if you pick the right one and tailor your prompts accordingly.
3. Experiment and iterate
You know the old saying: if at first you don’t succeed, try, try again. If you don’t get the perfect result the first time, try rephrasing your prompt, giving more context, or adding additional parameters. A simple “Could you make it sound more casual?” or “This is a blog post aimed at finance experts” could make all the difference. And if that still doesn’t hit the nail on the head: try, try again! The great thing about Gen AI is that it will never get tired of your constant tweaks and iterations.
4. Get creative, but not too creative
Gen AI is ‘smart’, but it still has a long way to go to compete with human intuition, abstract thinking and deductive reasoning. Being somewhat creative with your prompts may result in more creative, unique responses, but too far out of left field and you might end up with something that’s more obscure than useful.
For instance, if you prompted your Gen AI tool to “Describe our new product launch event as if it were a mythical epic,” you might get a response like, “In the grand halls of industry, the mighty product emerged from the mist, greeted by the cheers of the enchanted crowd”. While entertaining, this response isn’t practical for informing your audience about the event.
Remember you’re interacting with a programmed tool, not a superhuman mind-reader (or even a regular human). Write prompts that are coherent and structured, and you’re more likely to get lucid, accurate responses.
5. Be ethical
The ethical considerations around Gen AI usually focus on appropriate use of generated content, specifically relating to authorship, accreditation and plagiarism. But even before that, it goes without saying that your Gen AI prompts should follow the same ethical standards as any other writing or data input. Avoid requests that aim to provoke a biased, offensive or otherwise harmful result, which Gen AI may not be sufficiently programmed to avoid.
6. For bonus points, be nice
It’s still open for debate whether being ‘kind’ to Gen AI really matters. On the one hand, one expert has said “There’s nothing to suggest ChatGPT would provide different responses based on how politely someone frames a question”. On the other hand, limited user testing suggests that a polite request may yield a response that is more carefully formatted and friendly in tone.
Either way, there’s no harm assuming the tone of your prompt may help shape the tone of the response. As we’ve already discussed – and as happens to be the case ‘in real life’ too – a clear, detailed question is more likely to yield a lucid, on-topic answer. A prompt written in a polite, professional tone may elicit a polite, professional response, while deliberately aggressive or insensitive phrasing may result in copy that you wouldn’t want to use in a professional context (if at all).
Furthermore, if you consider that every interaction with Gen AI feeds into the ever-evolving language learning model – training it how to give responses for other users, not just you – there’s even more reason to be professional and respectful when writing your prompts. We probably don’t have to worry about the technology eventually becoming sentient and remembering who was nice to it, but we might all do our part for the quality and credibility of the tool in the long run. Also, it’s nice to be nice.
Before you get started
Some organisations have taken steps to limit or ban the use of Gen AI at work, but even if your workplace has openly embraced Gen AI it may have also implemented a policy governing its use. These policies can include restrictions on using Gen AI to create copy that will be attributed to your organisation, instructions on how to clearly notify readers that you have used Gen AI and guidelines for protecting confidential information. Remember to check if your organisation has such a policy, and familiarise yourself with its contents – even seek approval, if required – before setting out on your Gen AI journey.
