The ongoing evolution of effective AI prompt engineering

Just as artificial intelligence (AI) has undergone a period of rapid development and innovation in recent years, so too has the allied field of prompt engineering.

Designed to enhance the outputs delivered by AI tools, prompt engineering is an approach users can take to enhance the commands or questions they pose. The objective is to make the prompt as clear and detailed as possible to ensure optimal results.

When large language models (LLMs) first captured widespread attention in late 2022, so-called ‘zero-shot’ prompting was the norm. This involved users asking a specific question of an LLM and rapidly receiving the answer.

During 2023, prompting evolved as people began to ask multiple questions or request that AI outputs be further enhanced in some way.

During that year, role-based prompting also emerged. This involved a user asking an LLM to respond as though it was playing a particular role. For example, a user might prompt: “You are a finance professional and have been asked to create a detailed report from the attached spreadsheet” or “Act like an IT professional and explain the best way to implement a cybersecurity strategy compliant with government regulations”.

Last year, chain-of-thought prompting became more widespread. This involves a user asking an AI tool to undertake a series of steps to create an output. This has continued to evolve during 2025 and users are increasingly aware of the need for accurate prompting to achieve the best results.

Indeed, AI conversational designers are now being replaced with prompt engineers. With this simplistic approach, we’re replacing our developers and lines of code with subject matter experts developing in Microsoft Word.

The components of an AI prompt

There are six key components required to create an effective AI prompt. They are:

1. The task: Clearly state what you want the AI tool to do. This could be ‘write’, ‘summarise’, ‘compare’ or ‘translate’.
2. The format: Define the desired output format. This might be bullet points, a table, computer code or written text.
3. The tone or style: Nominate the style required, such as ‘formal’, ‘simple’ or ‘persuasive’.
4. Context: Provide relevant background information that the AI tool will need to understand the request.
5. Persona or role: Specify the role the AI should assume. This might be a marketer, a software engineer or a teacher.
6. Constraints: Add any limits or requirements that apply. This could be a word count, or whether stats should be included.

Additional prompt techniques

As well as these six components, there are a number of other techniques that can be used to make AI prompts even more effective. They ensure that desired outputs are returned as the highest quality and in the fastest time.

One example is the concept of ‘negative prompting’. This requires the user to explicitly state what they do NOT want included in the response. This avoids undesired outputs or off-topic content.

Another is the use of delimiters that separate parts of data or text. Examples include **** or #### and they will improve clarity for the AI tool and reduce confusion, especially when it comes to multi-part prompts.

Best practices for AI prompting

Following best practices when it comes to prompting will significantly improve the outputs that are achieved. One of the most important is ensuring prompt clarity. The AI tool will perform better if it is clear exactly what the user requires. Vague prompts are unlikely to deliver quality results.

The tone of a prompt is also important as it should be consistent, appropriate, and easy to understand. Prompts should also be complete and include as much detail as possible to guide the activity of the AI tool.

An essential skill

As AI technologies continue to advance, effective prompting has become an essential skill for anyone seeking to extract maximum value from these tools. The evolution from zero-shot to role-based and chain-of-thought prompting demonstrates how user techniques have matured in parallel with AI capabilities.

By understanding the six key components of a strong prompt – and applying additional methods such as negative prompting and the use of delimiters — users can significantly improve the accuracy, relevance and usefulness of AI-generated outputs. In many ways, prompt engineering has become a collaborative process between human and machine, where carefully structured instructions enable AI to deliver its full potential. As the field continues to evolve, those who master the art of precise and thoughtful prompting will find themselves better equipped to harness AI as a powerful partner in problem-solving and innovation. *Paul Rilstone is Vice President ANZ at Kore.ai.

Top image credit: iStock.com/narvo vexar