You have probably heard AI chatbots have the potential to revolutionise your business operations if you apply good AI prompt writing.
It’s true, AI prompt writing can do a lot for your business. The output from generative AI chatbots can help you:
Effortlessly create all kinds of content, from marketing plans to product descriptions to technical proposals
Analyse data and make recommendations
Streamline, automate, and enhance business processes
Inspire innovative new ideas
But how?
Generative AI is a type of artificial intelligence that creates (generates) new content like text, images, code, music, or videos. Generative AI models are trained on huge amounts of data so they can learn patterns, make predictions, and generate content that mimics human creativity and imagination. Large language models (LLMs) are designed to understand and generate text. And other advanced AI models can generate images and other forms of media.
Maybe you’ve already played around with AI prompt writing using chatbots like ChatGPT. Maybe you’re a little disappointed with their output. I know I have been. Sometimes it’s given me rubbish information that didn’t answer my question. Sometimes it even makes things up! But I never stopped to think that maybe the problem was my AI prompt writing rather than the chatbot itself.
As they say,
Garbage in, garbage out.
To get quality results from AI chatbots, your questions and/or instructions (AI prompt writing) need to be clear and descriptive while remaining concise. This practice is known as prompt engineering.
Importance of Well-Written Prompts
The responses from AI chatbots aren’t the same as old-school search engine responses – a bunch of links you need to click on and explore for yourself. Similarly, writing an AI prompt isn’t the same as typing a few words into a search engine.
Good prompts matter because they provide the AI with:
Clarity and direction – Prompts are the communication bridge between you and the AI. A well-written prompt gives the AI clear instructions, so it understands how to generate the specific content you want.
Quality and relevance – The more precise your AI prompt writing, the more accurate and relevant the AI’s output. This helps ensure it generates valuable insights and content for your business.
Efficiency – Starting your query with a well-crafted prompt saves you time and resources by reducing the need for multiple iterations and corrections.
AI Prompt Writing Pitfalls to Avoid
There are several hallmarks of a bad prompt that’s going to result in a bad response.
To start with, try to avoid the following when writing your AI prompts:
Vagueness – If your AI prompt writing is unclear or ambiguous, you’re likely to receive a generic or irrelevant response. For example, if you ask the AI, “How can I improve my business operations,” you probably won’t get any innovative new ideas or suggestions tailored to your specific business. Try including in your prompt some information about your business and what you want to achieve.
Lack of context – If you don’t give the AI chatbot sufficient context, it probably won’t understand what you want. Instead, setting the scene and including relevant details will help the AI generate output that’s relevant to your situation. For example, if you need to write a business plan you could give it a sample of another business plan and tell it about your target clients.
Complex instructions – LLMs are sophisticated, but overly verbose prompts might confuse them so they can’t interpret your query well enough to give you your desired results. Some LLMs have trouble understanding long pieces of information. You might get a better result if you break your prompt into chunks and work with the AI iteratively. Think of a well-constructed prompt as a series of building blocks.
Guide for Good AI Prompt Writing
1. Define your objective
The first step is to clearly identify your goal.
Do you need creative marketing content, a business proposal, data analysis, or a sales presentation?
For example, an Australian travel agency might want to create descriptions of tours of the Australian outback for different customer segments.
2. Structure your prompt
A good prompt isn’t a jumble of incoherent text. Good prompts are structured. As I said earlier, good prompts consist of a set of building blocks.
Here’s an example of how you could structure your prompt:
Instruction – Start with a clear directive for the task you want the AI to complete. For example, “Create descriptions of tours of the Australian outback…”
Context – Provide the AI with relevant background information. You should include relevant details such as descriptions of the target audience or the desired tone of the output. For example, “Customer segment 1 are retirees who like to travel in comfort but are active and curious. Address them in a friendly tone.”
Constraints – Set boundaries or limitations to guide the AI. For example, “Limit the tour descriptions to 500 words.”
Example – Include a sample to show the AI what you expect from its output. For example, you could include a description of an organised tour you used in a previous campaign.
3. Be specific
The more specific your prompt, the better. You can include keywords, key phrases, and relevant facts. For a travel agency, the prompt might include itineraries, locations, or target customer details.
4. Don’t be afraid to experiment and iterate
Using generative AI is a creative process and you might not get the right results with your first prompt. Try out different approaches, learn from the results, and then refine your AI prompt writing accordingly.
5. Consider prompt length
While detail is important in a good prompt, you should write your prompts so they’re as concise and focused as possible.
You might need to break a complicated prompt into subtasks so that the AI understands it.
Advanced AI Prompt Writing and Other Considerations
Once you’ve got the hang of AI prompt writing, try out some of these more advanced prompt engineering techniques:
Conditional prompting – Give the AI additional conditions or scenarios to consider. Going back to our travel agency example, you could tell it, “If the target customers are retirees, emphasise comfort and luxury.”
Negative prompting – Tell the AI what to avoid. For example, “Don’t use slang in customer communications.”
Prompt chaining – Guide the AI through a series of tasks, which may be subtasks of a larger, more complicated task. Each prompt builds on the response to the previous prompt, allowing for more complex and structured content to be generated.
Don’t forget, you’re responsible for the AI’s output and how it’s used. So, when working with AI, be aware of potential biases, limitations, and ethical considerations.
Data bias – Generative AI chatbots can give you biased responses they picked up from their training data. Review the responses and consider using some negative prompting like, “Don’t favour any gender…”
Copyright and plagiarism – The onus is on you, so always check the AI’s output is original and doesn’t infringe on someone else’s work. To mitigate copyright issues, ask the AI to only use reliable sources and to cite those sources. You can also run its response through a plagiarism checker.
Privacy and security – Always protect sensitive business data when using AI chatbots, whether it’s your data, your customers’, or even your competitors’ data. Unless the AI chatbot is a company-approved model sitting behind a corporate firewall, don’t input this kind of data into an AI prompt.
The Australian Government has created a Voluntary AI Safety Standard to guide Australian businesses on safe and secure ways to use AI.
