Wharton@Work

June 2025 | 

Supercharging Innovation with AI

Supercharging Innovation with AI

In today’s race to innovate, a new force has emerged — one that’s tilting the playing field in favor of those who know how to wield it. Generative AI is changing how organizations conceive, shape, and select new ideas. And according to Wharton professor Christian Terwiesch, co-creator (with colleague Karl Ulrich) of the Innovation Tournament framework, the impact is nothing short of transformative.

“Innovation Tournaments are built on a simple premise,” Terwiesch explains. “If you want better ideas, start by generating more ideas. Then, use structured evaluation to identify the best.” That formula — many ideas, winnowed through smart filters — is a proven method for managing innovation inside companies. What’s changing now is not only how ideas are generated, but also how they’re evaluated — at scale, with speed, and often with surprising sophistication.

A Machine That Prints $100 Ideas

Terwiesch vividly recalls the moment he realized the potential of generative AI. In late 2022, while advising a major bank on attracting young families and children, he decided to test ChatGPT (then version 3.5) as a creative collaborator. The prompt was simple: “You are a super-creative product executive. Generate 10 new banking product ideas for families and kids.” The AI’s results — products like Financially Fit, Smart Start, and Entrepreneur Express — were virtually indistinguishable in quality from ideas generated by the bank’s marketing and innovation teams.

One idea, Virtual Piggy Bank, even received a Shark Tank–style pitch from the AI. “Honestly, it was one of the better pitches I’ve seen — and I’ve seen thousands,” says Terwiesch. More importantly, the AI did this in seconds, at essentially no cost. “In most organizations, generating a single idea can cost $100 or more. What we have now is a machine that prints $100 bills.”

Putting AI to the Test

To measure how AI stacks up against humans, Terwiesch and Ulrich, both longtime instructors of innovation at Wharton, ran a side-by-side test using a favorite classroom exercise: generate a product for college students that costs less than $50. They randomly selected 200 student ideas from past classes and had ChatGPT-4 generate 200 more — half with a basic prompt, half using “one-shot prompting” (giving the model a strong example first).

To evaluate them, hundreds of college students rated the ideas based on purchase intent. The AI-generated ideas outperformed the human ones. The average purchase intent jumped to 49 percent with one-shot prompting, compared to 40 percent for the student ideas.

But the real difference came at the top. “In innovation, we care most about the best ideas — the outliers,” Terwiesch notes. Of the top 10 percent of ideas, 35 came from AI and just five from students. “That changes everything,” he says.

The Implications for Innovation Tournaments

For companies running Innovation Tournaments, the implications are clear. Generative AI is not a replacement for human creativity — but it is an accelerant. “At this point, leaving AI out of your innovation process means missing out on speed, scale, and quality,” says Terwiesch. “It’s quickly becoming a standard part of the toolkit.”

By dramatically lowering the cost and increasing the volume of idea generation, AI gives organizations a deeper pool of candidates from which to choose. That means a better chance of finding the hidden gems — the breakthrough ideas that drive real growth. The tournament model, long proven to help organizations filter for the best ideas, becomes even more powerful when paired with AI at the front end.

Smarter Evaluation at Scale

Generating hundreds of ideas is just the first step in the Innovation Tournament model. The real challenge? Figuring out which of those ideas is worth pursuing.

“Coming up with ideas is fun,” Terwiesch says. “Evaluating them is where things get tricky.”

Traditionally, evaluating early-stage ideas requires costly testing — talking to customers, building prototypes, and putting products into the market. Few organizations can afford to do that for dozens of ideas. But what if AI could take on that work too?

It turns out, it already can. Not perfectly. Not entirely. But powerfully — and increasingly.

Beyond Gut Instinct: Using AI to Evaluate Ideas

Once you’ve used AI to generate ideas, the next challenge is figuring out which ones are worth pursuing. Traditional methods — customer interviews, prototypes, and market tests — are expensive and slow. That’s where AI can step in again, offering new ways to stress test, visualize, and even simulate the innovation process itself.

The most obvious place to start is simply asking AI to score an idea. Will people buy it? Is the concept viable? Terwiesch and his team tried just that — asking ChatGPT to predict purchase intent for hundreds of ideas and comparing those scores to actual survey responses from college students. The result? Not great.

“We found that AI wasn’t much better than humans,” Terwiesch says. “But that’s telling. Experts are also terrible at predicting what people will buy.” The takeaway: don’t lean on AI for ratings alone — but don’t stop there either.

Where AI really shines is in stress-testing ideas. Instead of asking for a verdict, ask for concerns. Prompt it with questions like, “What are five reasons this product might fail?” or “What assumptions am I making that may not hold?” These kinds of queries surface potential blind spots and risks — quickly, cheaply, and often creatively. “It’s fast, it’s free, and it doesn’t require anything beyond smart prompting,” says Terwiesch.

AI also accelerates prototyping. What once required designers and engineers can now be done in minutes using image-generation tools like MidJourney or ChatGPT’s built-in visuals. Terwiesch demonstrates this with a hypothetical example: a fictional iPhone-powered bike pump. “I’m not an engineer,” he says. “But I can still mock up the idea, explore the physics, and iterate with AI’s help.” Whether it’s a sketch of a product or a storyboard of a service, visualization helps align teams and turn abstract ideas into something concrete — without ever leaving your desk.

Then there’s what Terwiesch calls one of the most promising frontiers: synthetic consumers. AI can simulate market feedback by roleplaying different types of users. Want to know how a 7-year-old with a $5 allowance might react to a digital piggy bank? Just ask. You can simulate dozens — or thousands — of consumers in minutes. “It’s like doing market research in the metaverse,” Terwiesch says. While still early days, the use of synthetic consumers is already being explored by political campaigns and forward-leaning marketing teams.

And if AI can generate ideas, prototype them, and simulate customers — why not go further?

Terwiesch envisions a not-so-distant future in which AI agents handle the entire innovation cycle: identifying needs, generating and refining solutions, testing concepts, and ranking the best options. “We’re not quite there yet,” he says. “But we’re getting close. And when that happens, innovation becomes even more scalable and continuous.”

In this future, AI doesn’t just assist innovation. It does innovation.

Redefining the Innovation Tournament

Innovation Tournaments were designed for a world in which human effort was the bottleneck. But with generative AI, the bottleneck moves. Now the challenge is deciding how best to use these new capabilities.

“Think of AI as an idea multiplier and a testing assistant,” says Terwiesch. “It’s not magic. It’s leverage.”

With that leverage, organizations can run bigger, faster, smarter tournaments. They can explore more paths, test more assumptions, and surface better ideas — with a fraction of the time and cost.

The rules of innovation haven’t changed. But the tools have. And as Terwiesch likes to remind his students: “The AI you’re using today is the least capable you’ll ever use.”

The smart move? Start using it now.