5 Ways We're Using AI at Work

This phase of the AI revolution feels personal. Many IDEO designers are artists or writers—crafts that generative AI is uniquely suited to disrupt. So, we’re getting ahead of it and embracing these tools to explore how our creative process might coevolve with AI.

As we would do with any new colleague, we hope to learn how to have constructive relationships with these technologies, understand them better, and collaborate.

Way back in 2019—before generative AI was everywhere—we created a deck of AI Ethics cards, which you can download here. It was a good start, but there is much work yet to be done on creating an ethical human-AI ecosystem that endures long after the hype cycle ends.

Here are a few of the ways we’ve been experimenting so far:

1. Research Synthesis 

The classic image of design thinking is people huddled around a Post-It–littered foam core board. During the pandemic, this process shifted to digital tools like Miro, Reduct, and Figjam.

Interaction Designer Takashi Wickes is using generative AI to further streamline the synthesis process—a method of organizing and interpreting research data in order to identify patterns, themes, and insights that can be used to inform the design of a product or service or experience. Using Reduct to transcribe his interviews, he highlights stand-out quotes. He then runs the quotes through Notion AI with a prompt like “What are the top 10 key takeaways from this conversation?”

Knowing that the AI doesn’t have the larger context of the project at hand, Takashi and his teammates use Notion AI’s output as supplemental input. Treating Notion AI’s output as a base to build on, they combine AI and human intelligence to go deeper, faster.

And, we’re already seeing startups form around this type of process: check out Vowel.

2. Concepting & Ideation

Designers come up with a lot of ideas. A typical ideation session begins with a prompt, something like: In a world oversaturated with screens, how might we support parents and children in connecting over the dinner table through games? We scratch our heads, think it over, and begin sketching out sacrificial concepts.

Recently, we’ve been experimenting with flipping this process on its head. Rather than coming up with an idea straight away, we use AI-based image generators like DALL-E or Midjourney for inspiration.

For a project on new dinnertime games, we fed DALL-E the prompt: “a family playing with their food during dinner.” The images rendered spark new directions that we might not have come to on our own. Mom with a carrot as a nose? Why didn’t we think of that? When you’re all at the table, why not use food as costumes in gameplay? Magical game design is all about turning expectations on their head. Now, we start to wonder… What other foods would make fun play pieces?

And of course, there are a cohort of new companies who are formalizing this process into a consumer tool: check out Fermat.

3. Critique

One thing we’re always exploring in our very riffy culture is how to get better at critique. There’s always this one thing to pick on: Weak insights. Insights at IDEO are concise statements that reflect our understanding about the current state of affairs in a given challenge and the way they are framed can be foundational to how we design solutions.

A few years back, before AI tools were everywhere, we made an insight analyzer chatbot. It looked for overused terms and poor word choice and would essentially chew you out. It was just an NLP bot that used natural language processing to match user chats with a set of known statements. The interface was a physical telephone (hotline). A funny gimmick with a simple learning: bots can get away with being mean.

Late last year, a team of brand and interaction designers started packaging up our favorite simple methods of critique and revisited the old insights hotline. Anne Graziano and I fed simple prompts such as “Give a brutally honest critique of the following insight: Meet people where they are at” to ChatGPT, and as might be expected, if you talk to an AI modeled after what’s said on the internet  it’s going to be very good at criticizing things.

We tried the OpenAI api (GPT3_davinci-003) and got decent results, so we wired it up to an emoji reaction in a Slack channel. When you add a “crit” emoji to an insight, it replies with a critique. The original prompt was harsh and getting a bit obstinate, so our group adjusted to the prompt “Give constructive feedback on the following insight:” which changed the tone and made it more creative and fun.

Our colleague Takashi figured out that the AI was pleased to critique quotes from classic movies and romcoms, which Takashi sourced from ChatGPT (prompt: “Give me the 10 cheesiest famous lines from movies”). He then fed the AI old chestnuts like, "Life is not measured by the number of breaths we take, but by the moments that take our breath away." (Hitch, 2005).

4. Marketing Copy

As community builders at IDEO, we host a lot of events. Recently, when writing up the copy for an internal informational session about the fundamentals and history of generative AI, I turned to ChatGPT to write the marketing copy. The output was pretty spot-on, and ended up informing both the content of the event and the copy we used to pitch it. We built on it, of course—inserting things that we thought were especially important for our community to learn—but it was a true collaboration.

5. Slide Decks

Google Slides are a staple of IDEO’s creative process. We use slides for project deliverables, ideation, business proposals, and internal events. We've found Tome.app, which creates AI-generated slides, to be a helpful springboard. Business designers are experimenting with the technology to refine their communications, interaction designers use it to think through how a new product might look, act, and feel, and our Play Lab is using it to explore new rules and play styles for games that do not yet exist.

Companies are forming around key parts of our creative process—capabilities which humans (and a few animals) once claimed as their sole domain.

In a 1948 documentary about U.S. agriculture called "The Land" the disruptive force of large-scale mechanization is described in a way that feels like it could be about this AI moment:

"These miraculous machines!
Do we shape them
Or do they shape us?
Or reshape us from our decent, far designs?

But we are learning.
We are learning to build for the future
From the ground up."

We’re curious: How are you using generative AI tools at work? We’d love to hear from you! Send us a note: Savannah Kunovsky & Danny Deruntz