By Contributor,Sarah Elk

Copyright forbes

AI won’t spark real innovation until we stop using it to validate ideas and start using it to break them.

Recently, OpenAI and Anthropic revealed how most of us are using generative AI at work. Anthropic reported that 44% of Claude’s business and developer traffic maps to computer or mathematical tasks, like code generation. These users are also more likely than others to automate directive tasks, chasing productivity gains. Of work-related ChatGPT messages, 40% are writing tasks, 24% are requests for practical guidance, and 13.5% are seeking information.

Useful, certainly, but not quite transformative.

At the same time, innovation has never been more urgent, nor more difficult, to get right. Companies are spending immense organizational energy on hackathons, design sprints, idea challenges, and the prioritization of tens and hundreds of use cases. Backlogs are full, yet few ideas make it to market. And when they do, the results can still be discouraging: Research shows only 5% to 25% of new products and services succeed.

AI holds the potential to change that. But we won’t get there if we continue to use it as a light brainstorm aid, asking it for email drafts, catchy taglines, or prototype images. For real breakthroughs, we need to aim AI at the right part of the innovation process.

When efficiency isn’t enough

My colleagues recently surveyed and interviewed a selection of leaders from Fast Company’s 50 Most Innovative Companies and found that these firms are harnessing the power of AI for creativity. In particular, they are using it to scale their innovation ambition and sharpen their understanding of customers.

The velocity is impressive: 31% of top innovators have cut their design-to-launch timelines by more than 20%, and 82% expect even more compression in the next five years.

AI can also reduce the odds of failure, through analysis of historical data, customer sentiment, and market conditions. In fact, 88% of innovation leaders say AI has already increased their innovation success.

MORE FOR YOU

Those wins matter. But they also reveal a blind spot. Most companies still point AI inward, optimizing processes instead of solving customer problems or building new, nonobvious businesses to meet new, nonobvious needs. To really harness AI’s power, we have to shift from convergent thinking to divergent thinking.

Divergence on demand

Currently, most AI use cases support the idea-narrowing, or “convergent,” phase of design thinking—tasks like synthesizing research, forecasting trends, clustering concepts, and analyzing options. That’s valuable but limiting.

Large language models were built to predict the most likely next word, the correct pattern, the most expected response to avoid hallucinations. So, it made sense for them to generate safer, predictable outputs.

But now, AI models are maturing. As they improve, the nature of the prompt will shift from “Find the next best answer” to “Explode the space of possible answers.”

What happens when leaders start using AI for the idea-generating, or “divergent,” side of design thinking? What if they intentionally pursue difference over consensus?

AI’s role in innovation won’t just be about moving faster but also going broader and deeper. Divergent use cases will be proactive, not reactive. Leaders will push AI to articulate latent needs and surface market-producing insights. They will aim higher than search, hypothesis validation, and prototype generation to explore wildly different ideas that can bloom into new products, services, and businesses.

How to prompt true innovation

Leaders can shift toward divergent thinking by changing how they engage AI. Here are five tactics to start with:

Use open-ended prompts. Don’t overly engineer inputs. More open-ended requests can open more doors. For instance, ChatGPT itself suggests starting with something like “Ask for five paths that would terrify an incumbent—and why.”

Generate contradiction. Ask AI to pair ideas that can’t both be true, then explore the tension. Or request “wrong” answers and try to uncover hidden assumptions.

Push evaluative boundaries. Ask AI for ideas in categories such as “radical,” “incremental,” “tech-driven,” or “low-cost” to avoid obvious solutions.

Change perspectives. Prompt AI to take on the persona of a competitor, a customer from an unexplored segment, a futurist, a skeptic, a historical figure, or perhaps a leader from a different industry to spark new ways of looking at old problems.

Iterate. Play devil’s advocate or build on initial results to create new ideas. For example, after an initial brainstorm, prompt the chatbot to “combine ideas #3 and #7 into a new hybrid concept and evaluate market fit for [audience].”

Finally, the key to success will be balance between man and machine. AI can manufacture more “what ifs” per minute. But human judgment and creativity will still be essential to determine which provocations unlock genuine value, which tensions are worth exploring, which risks deserve a bet. What’s more, the partnership only works if people have an open mindset. While many AI-generated ideas will be discarded, of course, the real discipline will be differentiating bad ideas from our own closed-mindedness. AI can spark inspiration, but humans still need to supply the intuition, empathy, and an experimental mindset to take bold, imaginative leaps.

Editorial StandardsReprints & Permissions