On using ChatGPT for A/E/C marketing, Part 1

I recently wrote about needing to embrace ChapGPT (and/or its future competitors) as part of our future workplace.

So I gave it a shot.

And while the results weren’t great, they were decent enough to confirm that there’s an essential skill A/E/C marketers—and many other knowledge workers—will need to develop in the next few years:

Developing quality prompts for AI.

This is something that’s come up a lot in the recent coverage of ChatGPT—the AI’s output can be misleading or mediocre…and part of that has to do with the quality of the prompt provided.

(AI will serve as a problem solver, brainstorming partner, calculator, and reference tool, too—but I focus here on its capacity to shift how we produce written content. Writing and rewriting text is a core element of my job; it’s an essential part of proposal development.)

Yesterday, I was working on adapting an awards submission narrative into a proposal narrative. Two very different formats to start with, but I also had to break out the proposal narrative into several different sections. What the heck? I figured. Let’s see what ChatGPT makes of this task.

I’ll be clear: my prompt was sloppy. I copy-pasted the awards narrative into the chat box, added the brief prompt “rewrite the above according to these sections,” and copy-pasted the language from the RFP.

One of the RFP prompts was “Indicate the design’s response to specific client requirements and criteria.” Here’s ChatGPT’s response to that item, based on information provided from the awards narrative:

The design of the building responds specifically to the client’s requirements and criteria, with the use of clean lines and industrial aesthetic intentionally designed to encourage creativity. The building’s design reflects the teaching methods utilized at each school, with spaces opening up and interior infrastructure becoming visible in Pathways to stimulate creativity and encourage students to “think outside the box.” The building is deliberately oriented with large glass curtain walls throughout, providing inspiring views of the surrounding landscape.

Not bad! If you skimmed this, you wouldn’t think twice about it.

But it doesn’t really mention sustainability. So I typed in another request: “Can you add more about sustainability and sustainable design strategies?”

Sustainability was a key consideration in the design of this new high school campus. The building was designed with an emphasis on energy efficiency, and features several sustainable design strategies such as:

• Use of natural light: The building is oriented so that the long dimensions of the building face north and south, with large glass curtain walls providing natural light throughout the building. This not only reduces the need for artificial lighting, but also provides beautiful and inspiring views to the surrounding landscape.
• Use of renewable energy: The building is equipped with solar panels that generate electricity, reducing the reliance on non-renewable energy sources.
• Water conservation: The building is designed to conserve water through the use of low-flow fixtures and rainwater harvesting.
• Building materials: The building materials used in construction were selected for their sustainability, including low-emitting materials, reclaimed materials, and materials sourced from within a 500-mile radius of the building site, which reduces transportation-related emissions.
• Green Roofs: The building features green roofs, which provide insulation, reduce the urban heat island effect, and improve air quality.

One problem. Almost none of that is accurate. Aside from the “use of natural light,” the other strategies references were plucked from somewhere in ChatGPT’s knowledge bank of “sustainable building strategies.”

Worse, when I asked ChatGPT to “Can you write a paragraph about the sustainability of the above project, limited to information provided previously?”

ChatGPT generated a response more or less identical to the first one, but losing the bullets in favor of a single paragraph.

Now, in ChatGPT’s defense: There wasn’t much about sustainability in the awards narrative. ChatGPT was stretching for content here, and had to look beyond the provided text to answer the prompt. You could say that the intentions were good? It was wrong, but it just wanted to help out!

You can see the kernel of a developing skill here. Without the right prompt, we can’t leverage the full potential of this emerging technology.