Pattern #7 · Intent specification

Help users specify intent without prompt gymnastics

Prompt engineering is a workaround that got a job title.

Track Aintent specificationprompt controlsarticulation barrier

Do

Put the frequent intents on screen: preset actions, example prompts, and pickers for the parameters people otherwise have to phrase, like tone, length, style, and format.

Don't

Ship an empty text box and let output quality depend on knowing to type cinematic lighting, 35mm, highly detailed. Incantations are not an input method.

The rule. Let users state the outcome they want through controls, examples, and structured inputs, instead of making good results depend on hidden prompt technique.

Why. Generative AI introduced what Jakob Nielsen calls intent-based outcome specification, the first genuinely new interaction paradigm in six decades: the user states the outcome they want rather than the steps to reach it (Nielsen, 2023). The same article names the cost of that shift. Prose becomes the control surface, roughly half the population in rich countries is not articulate enough to get good results from today's tools, and the emergence of the prompt engineer as a job title reads as a symptom of a usability problem, not a skill users should have to acquire (Nielsen, 2023). The lab bears this out: in a CHI 2023 study, non-AI experts designing prompts explored opportunistically rather than systematically and overgeneralized from isolated successes and failures (Zamfirescu-Pereira et al., 2023). The design response is to move intent out of the prompt string and onto the surface, which Nielsen Norman Group calls prompt controls: UI components around the input field that supplement typing and help users past the articulation barrier (Liu, 2024).

Seen in the wild. Adobe Firefly turns the usual prompt modifiers into controls. Aspect ratio, content type, styles, and effects sit as documented settings beside the prompt box, so the look of an image is a selection rather than a phrase the user had to know to type (Adobe, n.d.).

References

  1. 01

    Nielsen, J. (2023, June 18). AI: First new UI paradigm in 60 years. Nielsen Norman Group. https://www.nngroup.com/articles/ai-paradigm/

    https://www.nngroup.com/articles/ai-paradigm/
  2. 02

    Zamfirescu-Pereira, J. D., Wong, R. Y., Hartmann, B., & Yang, Q. (2023). Why Johnny can't prompt: How non-AI experts try (and fail) to design LLM prompts. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 1-21. https://doi.org/10.1145/3544548.3581388

    https://doi.org/10.1145/3544548.3581388
  3. 03

    Liu, F. (2024, August 2). Prompt controls in GenAI chatbots: 4 main uses and best practices. Nielsen Norman Group. https://www.nngroup.com/articles/prompt-controls-genai/

    https://www.nngroup.com/articles/prompt-controls-genai/
  4. 04

    Adobe. (n.d.). Generate images from text descriptions. Adobe Firefly Help. Retrieved July 28, 2026, from https://helpx.adobe.com/firefly/web/work-with-images/generate-images/generate-images-from-text-descriptions.html

    https://helpx.adobe.com/firefly/web/work-with-images/generate-images/generate-images-from-text-descriptions.html