Prompt engineering is one of the must-have skills at this point, but let's start from scratch: what TF is prompt engineering?
Prompt engineering is the art of crafting clear instructions for AI. By carefully choosing words and providing examples, you can guide AI to generate exactly what you want, whether it's text, images, or other creative content.
That is more or less the idea, however, it's difficult to learn how to prompt. It often relates to the way a system works (e.g. Power Automate, prompting won't use natural language per se, but more of a Power Automate logic-based prompting), but in the deep core is related to one very basic skill: communication. As someone with ADHD, communicating is quite challenging sometimes in a weird but unique way. I assume everything in my mind is obvious for the rest and every thought I have requires no context or introduction for the people I'm speaking to. I didn't figure this out until not too long ago when some colleagues reminded me of the need for context over information.
It might seem unrelated but I can tell you there is a reason why I'm sharing such personal information. Prompting is communicating, having a clear structure of what you want, how you want, and what should be used to achieve your goal. Let's see below two different examples (promise it will make sense later).
One of the things that I've been using to master prompting engineering is Microsoft Designer. Designer provides many functionalities, but we will focus just on image creation :
Microsoft Designer is an AI-powered tool that simplifies graphic design. It offers a range of features to help users create professional-quality images and designs efficiently. A standout feature is the Image Creator, powered by DALL-E, which generates images based on text descriptions. This allows users to bring their creative visions to life quickly and easily.
As context, I created a new project called visionsofarchitecture (see examples at the bottom of the post) which helped me improve my prompting skills by crafting some - what I consider - very nice images that combine architecture, fabrics, and artificial intelligence. Keep reading for a couple of takeaways on what I've learned along the way and some examples of how Microsoft Designer can be used as your trainer for your prompting skills.
The main point for prompt engineering is understanding what the outcome you want to obtain, and describing as exhaustively and in detail as possible, this will improve the result you will get, and this principle is cross any platform or AI system you want to use.
Below you find an image that Microsoft created to help you craft the best prompts. It can be used as a guide.
You can also access official resources here:
Let's jump into some examples.
Example 1: Specific prompt and vague prompt
Specific prompt | Vague prompt |
Design a high-rise residential building with a focus on sustainable living. Incorporate rooftop gardens and communal spaces. | Make a tall building. |
As stupid as the above sounds, these two cases are either going to make or break the way we get the results. If you need proof, keep reading.
Specific prompt | Vague prompt |
![]() | ![]() |
There are plenty of factors that are important to obtain the best results. Sometimes you will need a simple answer, and sometimes you want to have a complete reply, with sources and references. It all depends on what you are aiming for.
Example 2: Elaborated prompts
Prompt 1 | Prompt 2 |
Design a dynamic Brutalist office tower featuring a precast concrete facade punctuated by oversized, recessed windows. The building should boast an open floor plan centered around a dramatic, 20-story high atrium adorned with a vibrant, living red wall and a canopy of skylights. Incorporate bold geometric shapes and contrasting colors in the facade, such as striking yellow accents or deep blue inlays. The tower should rise 20 stories above a 10,000-square-foot footprint, creating a powerful, sculptural presence on the skyline. | A 20-story, 10,000-square-foot Brutalist office tower is envisioned, its imposing form defined by a precast concrete facade punctuated by deep-set, oversized windows. The building's heart is a dramatic, 20-story high atrium, a vibrant space illuminated by a skylight canopy and featuring a striking, living red wall. The open floor plan is further accentuated by bold geometric elements and contrasting color accents, such as striking yellow and deep blue, integrated into the concrete facade. |
While the example is weird, it proves a point. See how two equal prompts land in two completely different results:
Prompt 1 | Prompt 2 |
![]() | ![]() |
As you can see above, even equal prompts will get different results. Reading both prompts, it's quite clear that the concept is the same. Is the order of the factors that alter the result. As prompts are crafted for you, with your unique communication skills and needs, the results obtained are likely to be unique as well. But this might not be the case always. A new experiment is coming - stay tuned.
Long story short, if you are not ready to start prompting in Copilot, Bing, Power Platform, or Dynamics in and for Copilot, this is an easy and fun way to improve your skillset and acquire a unique ability, to communicate with large language models and obtaining an outcome that is useful for you and your work or business, or maybe just for hobby - as I'm doing.
Commenti