When using AI, we do not simply want to see the results it produces. We also want to make sure that the final output actually aligns with what we need and expect. At its core, using AI is about being able to guide both the conversation and the results toward a desired outcome. This is what we commonly refer to as prompting.

A prompt is essentially a series of instructions, questions, or commands that we provide to an AI system to guide it toward producing the results we expect. Prompts can also be combined with images or audio, allowing the system to generate content based on the references we provide. Although prompting may sound simple, effective and meaningful prompting involves more than simply adding detailed specifications. It also involves guiding AI to engage in “role-playing” so that the resulting output can be more credible, relevant, and accurate.

For AI, role-playing is not merely a way to make interactions more engaging. It can also be a strategic approach to maximizing AI performance. Role-playing is particularly useful because it provides context, establishes boundaries, and “activates” a specific persona. This can transform AI from a simple answer-generating machine into a more adaptive tool capable of offering ideas and responses that are relevant to a particular situation.

Through a workshop recently held between Universitas Dian Nusantara (UNDIRA) and Paideia as part of the Google for Education and Gemini Academy programs, we explored a range of techniques for using AI more effectively and efficiently. Among the various approaches discussed was the PARTS prompting framework.

The PARTS framework stands for Persona, Aim, Recipient, Theme, and Structure. Persona refers to the role we assign to AI when asking it to develop or formulate an output. For example, we might instruct it, “You are a professional nutritionist.” Assigning a specific persona can help shape the AI’s language, perspective, and depth of response so that the output becomes more relevant to the particular profession or context.

Aim, or purpose, refers to the objective we want the AI to accomplish. Clearly stating the intended outcome helps the AI produce an output that is more focused and relevant to our needs.

Recipient refers to the intended audience or recipient of the information we ask AI to produce. For example, we might specify, “I am creating this educational material for first-year university students.” By identifying the target audience, we allow AI to adjust the tone, vocabulary, and complexity of the information accordingly.

Next, Theme refers to the desired style, tone, and other relevant parameters. In simple terms, what kind of output do Kawan UNDIRA want? Should it be humorous, formal, or perhaps semi-formal? Defining the theme helps ensure that the resulting content reflects the intended atmosphere and communication style.

Finally, Structure focuses on how the output should be organized and formatted. For example, we could instruct AI to “make it short, concise, and easy to digest and understand.” Providing structural guidance helps ensure that the final draft is not only informative but also presented in a format that suits its intended purpose.

When working with AI, we should not view it merely as a companion or a tool that provides instant answers. Instead, we should treat AI as a capable assistant to whom we give clear direction and responsibility for carrying out a task. The clearer the direction we provide, the more focused and organized the resulting output can become.

Mastering the PARTS framework and role-playing techniques can take our use of AI beyond simple question-and-answer interactions. We are no longer simply asking questions and accepting whatever answers AI provides. Instead, we are designing instructions, assigning roles, and directing AI much like we would when collaborating with a professional colleague.

At the same time, remember that applying PARTS and role-playing techniques should always go hand in hand with professionalism and integrity. Make it a habit to create structured prompts, review and verify every AI-generated response, and use AI as a thinking partner rather than a replacement for human reasoning.