Designing for the Public Sector with Generative AI

With a few rules of design, government leaders can harness new generative AI tools to help design the innovations of the future

By Alan Holden, Ben Szuhaj, Joe Mariani, and Tasha Austin

The explosion of generative AI tools like ChatGPT, MidJourney, Codex, and Dall-E is proving that AI can be successful in creative endeavors ranging from image creation to product design to document authorship.

But this does not mean that human designers are obsolete. AI offers the potential to exponentially scale the volume and diversity of innovative concepts. However, AI cannot yet understand the needs of users or the context of problems, so it needs humans to set parameters for that search or make decisions on which solutions best navigate an organization’s trade-offs. AI catalyzes human innovation rather than replaces it. So as these AI tools mature, public-sector leaders should consider the following:

What if the next wave of government innovation is powered by artificial intelligence?

The magic of generative AI isn’t in the silicon; it’s in the process—the way humans and AI work together to unearth new, valuable solutions to hard problems. That process is known as generative design.

Using generative design, a human designer can set up the process, input parameters, intervene partway through to tweak constraints, and, ultimately, select the best design or designs. The designer could then alter the designs personally or input them back into the process to be refined. The computer generates a variety of solutions and tests them against constraints; successful variations are amplified until multiple correct—and often unconventional—answers are produced that satisfy all constraints.

In many ways, generative design is the ultimate white-space innovation tool, enabling designers to exhaust entire solution spaces in a fraction of the time, freeing up the human to think strategically about where to play next. For example, when researchers applied the same kind of reinforcement learning used in generative AI to AI playing board games, it was able to uncover new strategies that had never before been seen in the thousand-year histories of games like chess and go.

For this and many other reasons, forward-thinking government leaders should begin considering how generative design could transform their operations, increase innovation, and deliver real value to the public.

Potential applications of generative design in the public sector include facility design, urban planning, policy creation, customer experience, learning and development, and reducing carbon emissions.

To set themselves up for success, potential adopters of generative design need to keep in mind a few key considerations:

  • Put the user at the center. Designers working in generative design should reflect the perspectives of end users in the prompts given to the AI and the decision criteria the designers use to select a final design.
  • Tap into the ecosystem for needed capabilities. Key partnerships with academia, technology providers, and nonprofits can help government agencies develop the skills and tools they need to use generative design effectively and ethically.
  • Implement strong governance. Significant care needs to be taken to embed ethical controls into every step of model development and use.
  • Don’t forget real-world constraints. Generative Design is a powerful tool, but even such a powerful tool can bump into limitations of what is possible in the real world.

Generative Design offers an opportunity to fundamentally transform how government interacts with citizens and stakeholders. But for it to become a reality, forward-looking government leaders must begin laying the foundation today.

Generating designs with generative AI

Get in touch

Tasha Austin
Principal | Deloitte & Touche LLP
Alan Holden
Senior Manager | Deloitte Consulting LLP

Fullwidth SCC. Do not delete! This box/component contains JavaScript that is needed on this page. This message will not be visible when page is activated.

Did you find this useful?