We Tried to Build a Health Venture With ChatGPT
< THINKING

We Tried to Build a Health Venture With ChatGPT

What worked—and what didn’t—in the process.
words:
Amina Jambo
Angela Kochoska
Tom Antony
Ziyuan ‘Zoey’ Zhu
visuals:
No items found.
read time:
4 minutes
published:
November
2023

Healthcare has a lot of big problems to solve. Could AI help? To answer that question, we teamed up with Healthworx Studio to run an intensive five-week sprint and design a business that would address issues of access in rural health, leaning on AI to supercharge the process.

We kicked off with a lot of questions: Could we produce more rigorous, higher-quality results if we used gen AI tools to augment our team? Would it come up with innovative ideas for how to make healthcare more equitable? Could we get further faster? The answer was yes—with a lot of caveats. (For a deep dive into our approach, principles, and learnings, check out our Designing with AI Report.)

It quickly became clear that while Gen AI tools can’t replace human ingenuity, they certainly fuel it. When it came to generating solutions to a problem this complex, AI had very little to offer in comparison to human teammates. But it did help us innovate much faster and with more depth, bringing us up to speed on initial stakeholder context and synthesizing and incorporating datasets —crucial parts of the process that can take a lot of time. Here’s where we succeeded and failed, and what that taught us about using AI going forward.

Prompt: An infographic with a dark background with a lot of imaginary logos in rows and columns, one logo is much bigger than the others, dark background, dominant teal, retro poster style

Opening the aperture

Did a machine just listen to dozens of podcasts, summarize them, and give us themes in an hour? Yes it did. Thanks to tools like Preplexity.ai, which enables social listening on platforms like Reddit, we were able to quickly synthesize diverse perspectives from varied sources and distill themes that seamlessly integrated into our research process. 

Working with these tools feels like being bitten by a radioactive spider and gaining a suite of superpowers. We can rapidly develop contextual awareness of numerous stakeholders and their pain points in relation to the challenges we aim to address. With our lens widened, the questions we asked became more pointed and the concepts we designed more crisp. “The AI enables efficient pattern matching across massive datasets, unlocking value from information that was previously siloed or inaccessible,” explains Jaime Goff, Product Design Lead at Healthworx Studio. “By leveraging these capabilities, we can incorporate a breadth of knowledge beyond what our team could internalize before. It really expands what we can achieve by connecting us to data in new ways.”

Prompt: Five people at separated cubicles working on computers, dark background, retro poster style

The importance of coming together

But as our use of AI tools increased, there were unexpected sandtraps, too. For one, we noticed a steep decline in the quality of our team interactions. We thought that getting an AI boost would be “like having an extra designer in the room with us.” But the reality was we each retreated into our corners with our own AI assistant, forging ahead with new ideas. When it was time to reconvene, we were on wildly divergent tracks, and re-aligning to move forward as a team was an enormous undertaking. 

Another caveat: While generative AI tools helped us gather new forms of data, it was still critical to hear directly from the people who would be using these solutions. 

Nicole lives on a ranch the size of the City of San Francisco. After her husband broke his leg in a freak accident, she had to make a harrowing journey, traveling two hours to the nearest hospital. Stories like Nicole’s taught us a lot about what it’s actually like to access healthcare in rural areas, and the resilience it requires. It was the people we spoke with who brought nuance to the problem our team was trying to solve, inspiring our most critical insights and building confidence in our concepts. As much as we tried, we could not find an artificial substitute that could come close.

It was the people we spoke with who brought nuance to the problem our team was trying to solve, inspiring our most critical insights and building confidence in our concepts. As much as we tried, we could not find an artificial substitute that could come close.

Smelling the AI B.S.

When reviewing a few AI-generated venture concepts with the Healthworx team, the room fell silent. The concepts were all variations of the same idea, repeated and rehashed. “Generative AI can rapidly spin up concepts that seem good enough at first glance,” Goff says. “But when you scrutinize them, the limitations become obvious.” Even after extensive prompting with our AI co-pilots, we were still back to the same place, with a bunch of repetitive concepts. 

But when the team mined our own insights, we were able to find a truly compelling venture idea.

Late into the night, when the team was riffing, we shifted our approach from working alongside AI to guiding it. We had ChatGPT review and improve on its own ideas, asking it to strengthen the overall concept as it went. The tool acted like jet fuel, helping us add rigor and build out the venture, ultimately producing a successful pitch presentation that included high-fidelity mockups, early working AI prototypes, and an in-depth pitch deck.

Prompt: A girl disentangling an electronic neural network algorithm, dark background, simple lines, retro poster style

Supercharging our process

We’re in the caveman era of generative AI, and there are plenty of unintended consequences to be concerned about. But we’re optimistic about how a human-AI partnership that leverages human creativity and nuanced subject matter expertise alongside AI tools allows us to quickly outsource parts of the process, add rigor, and iterate faster, supercharging design and innovation.

And while the AI hype is founded, its limitations, which became patently clear in the ways it addressed a complex challenge like rural healthcare, made us appreciate how important humans are to this equation. Goff says it best: “This project showed the value of individuals bringing their unique brains and nuance to the table to work alongside AI. Our minds provide the spark that drives real progress.”

For those who crave the details, learn more about our approach, principles, and experiments with our transformative collaboration with Healthworx Studio here or reach out to ai@ideo.com

Visuals created with Midjourney.

No items found.
Amina Jambo
Design Research Lead
Amina Jambo leads research on teams across a variety of industries, from healthcare to virtual reality. Prior to IDEO she used human-centered design strategies to develop delightful and innovative systems to increase habitability of extreme environments. Her background sits at the crossroads of architecture, physical science, and engineering.
Angela Kochoska
Senior Designer
Angela Kochoska a is a data scientist who leverages the power of data-driven design research and emerging technologies to fuel human-centered design. She works at the intersection of design and technology through projects where she often experiments and prototypes with Generative AI, natural language processing, and mixed methods analysis.
Tom Antony
IDEO Alum
Tom Antony is an alumnus of IDEO who served as a Senior Design Director with expertise in Venture, Product, Strategy, and Emerging Technologies (including Generative AI). He manages strategic and innovative projects across sectors with a particular focus on Tech and Health, creating a bridge between creativity and business rigor.
Ziyuan ‘Zoey’ Zhu
Software Designer
Zoey is a creative technologist and multidisciplinary designer at IDEO, she create tangible future through the lens of human-centered design, emerging technology and sustainability.
No items found.

Subscribe to the IDEO newsletter

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.