As artificial intelligence technology has been hyped over the past seventy years — and often failed to live up to expectations — the field has suffered through a number of “AI winters.”
But some of the latest advancements in neural networks and generative AI could mean that we have seen the last “AI winter,” write Paul Campbell and Stefan Bartschat in their new book, The Corporate Innovator’s Playbook. Instead, they worry we may be entering an “innovation winter.” In the excerpt below, they detail how innovation managers can avoid the deep freeze — and reposition their work for a new era.
Campbell is a global innovation leader who has built multi-billion-dollar businesses for companies like HP, Philips, Schneider Electric, and W.L. Gore. Bartschat has been a senior engineering executive at IBM and Electronic Arts, as well as COO of the wearable tech startup Lumo Bodytech. (On May 20th, Campbell will be our guest for a LinkedIn Live conversation about the book.)
• • •
Innovation managers took all of the recent AI advancements in stride at first. After all, the whole point of the managers’ existence was to scout new technologies, perform experiments, and nurture the most promising ones into products that create new revenue streams. If anything, the managers were anticipating more interest and resources from headquarters to engage with a new class of startups that had adapted the massive LLMs found in ChatGPT to new industry-specific applications.
But back at corporate headquarters, where leadership was already reeling from distractions related to COVID-19 shutdowns and supply chain distractions, ChatGPT raised alarm bells. It seemed more critical to ensure that R&D had the skills to stay up to speed with the rapidly evolving generative-AI landscape. Academics prognosticated that LLMs could be as influential on businesses as the shift from steam power to electricity early in the twentieth century. CEOs felt the need to show their boards, company shareholders, and their peers that they were “current” on this potentially giant disruption, and a common belief set in that not doing enough with GenAI could be a fireable offense.
Therefore, corporations founded AI centers of excellence, and since budgets are usually a zero-sum game, corporations diverted resources from innovation to fund them. AI winter has turned into innovation winter: Innovation centers started closing, either terminating staff or sending them home to toil away back in their old marketing or R&D departments.
But was this the right move? Are AI centers of excellence accomplishing anything? We don’t believe investing in AI centers of excellence at the expense of innovation centers is a great idea.
A center of excellence may enable the CEO to check the AI box, but it’s unlikely to generate value from new products or revenue streams.
First, we don’t see GenAI automatically creating large amounts of value across the enterprise. Large corporations optimize their staff to generate a continuous stream of incremental improvements; corporations aren’t as good at adopting management initiatives that involve big, disruptive change.
Hence, a center of excellence may enable the CEO to check the AI box, but it’s unlikely to generate value from new products or revenue streams.
Second, GenAI deployments require expertise in software and data, which are usually the domains of the corporate IT department. IT, in turn, is likely to invite external system integrators to pitch in on this effort, who will see this project as a chance to finally implement that long-postponed grand unifying data project. The end result? A complex scoping, evaluation, and testing approach that takes far too many resources and takes far too long—without any new products to show for it.
The multidisciplinary innovation center team has already demonstrated how to turn interesting technologies into corporate revenue streams, and the center’s partnership-focused approach is well suited to integrate the ever-evolving AI technologies into its opportunity funnel.
In contrast, the multidisciplinary innovation center team has already demonstrated how to turn interesting technologies into corporate revenue streams, and the center’s partnership-focused approach is well suited to integrate the ever-evolving AI technologies into its opportunity funnel. Therefore, the innovation center is best suited to drive AI technologies into the corporation. Of course, adding some AI skills to the team would be prudent to better assess outside partners and to gauge deployment opportunities for operations inside the company.
AI Broadens the Innovation Funnel
Generative AI is poised to dramatically change or even eliminate most business processes of a large corporation. It may also turn out to be the single most influential contribution to accelerating innovation in our lifetime. This is because GenAI excels at the tasks that innovators need most:
- rapid synthesis of large amounts of data (market, customer, internal ops)
- idea generation
- concept generation
- prototyping (primarily software for now)
- customer feedback and synthesis.
You can now do all those tasks rapidly and efficiently, virtually at the push of a button.
For example, you can use ChatGPT for brainstorming with simple input prompts, such as “Give me ten ways to improve our electric toaster.” And, unlike with human counterparts, brainstorming quality improves if you give ChatGPT feedback on its initial result, providing more ideas that help you home in on the most creative and effective ones.
GenAI tools can also help you validate these ideas, generating sample product photos (or even websites) that you can share with potential customers for feedback.
Of course, understanding the potential weaknesses of AI is important. First, we must realize that using generative AI is a two-way street. By interacting with the tool, the user is further contributing to its training data, which can be widely shared. That may be fine for advice on how to paint your child’s bedroom, but if some of the inputs are related to sensitive corporate data, creating a private instance of the tool you’re licensing is critical. This means that the tool’s learnings only influence your instance and won’t be passed back to make the general AI (and potential competitors using it) smarter.
Second, it’s helpful to realize that the principal design goal of generative AI is to repackage existing training data into new forms that best satisfy a particular prompt. This means that the neural network will favor ideas in the middle of the bell curve of acceptance — in other words, those that most people would find interesting. It would place less emphasis on the ideas at the outside of the bell curve, where consensus is limited and which most evaluators would likely judge as impossible or crazy. Yet the most effective ideas are often generated in those fringe areas of the bell curve!
Human innovators are still needed to prompt the AI to push it toward the edge, where the truly disruptive ideas lie hidden. And often, humans are still providing the “spark,” now aided by a virtual roomful of data within which they can make connections.
Third, if we use GenAI for market research, we need to recognize that it’s prone to occasional hallucinations: errors generated with confidence by the underlying models. Just as the innovation loop needs humans to create the spark, we’re also necessary to quality check the sources behind the research presented.
The final issue is not a weakness for AI, yet it still presents a challenge for an unprepared innovation manager: the pipe-jam problem. After adding the spark provided by its human counterparts, generative AI will make idea generation and initial validation much faster than conventional approaches. As a result, more and more promising ideas will enter the innovation funnel.
In order to fully evaluate an idea, the innovation team must build the idea into an MVP that they can sell in small volume. This takes time and effort, and with the flood of promising ideas entering the funnel, the innovation center may need to accelerate its processes to advance them. One way to do this is to somewhat relax this premise and add evidence building earlier in the process. For example, a team could build a faux product web page to see how eagerly potential customers push the Buy button. Another strategy involves hiring more outside partners to multiplex the design and engineering activities for more simultaneous projects.
On a related note, a key challenge for many corporate development efforts is to accept that an idea is no longer worth pursuing. Generative AI will only exacerbate this problem. Innovation managers need to get even better at canceling projects to keep the more promising ideas in the innovation funnel moving.
Process Innovation
This is an opportunity to expand the mission of the innovation center from a primary focus on product innovation to more dedicated time for process innovation — assisting in making internal processes smarter and more efficient. Note that innovation centers have always spent some time in this area, but in the past, it was difficult to get sustained interest from the internal teams to adopt new technologies. This has definitely changed: Innovation managers are fielding more and more internal requests to evaluate promising new ideas and technologies to make operations more efficient.
Given that successful process innovation requires a deep understanding of existing operations and specialties, we recommend that the innovation center form a collaboration: The team or function making the request should form its own dedicated subteam with support from (or active membership by) members from the innovation team. One goal with this collaborative approach is to separate projects that the team or function can do on its own from projects that require the special skills of the innovation center.
The innovation team’s main focus should remain on transformational and disruptive projects. Not every internal project requires this skill set, so establishing criteria for selection is important. Use the following criteria to perform the analysis:
- size of the payoff
- newness to the company or function
- problem complexity
- incremental versus transformational process.
Variety and Personalization
The net effect of integrating artificial intelligence into the innovation process (and operations) is that more ideas can be generated, developed, and commercialized more quickly and cheaply.
As AI tools make innovation faster and more efficient, business model disruptions are sure to follow.
This faster product-generation cycle will increase the variety of solutions available, accelerating the existing trend toward more-personalized products and experiences across all industries, from health care to automotive to consumer products. As AI tools make innovation faster and more efficient, business model disruptions are sure to follow. For example, if automobile models can be made in near-infinite varieties, perhaps a subscription model no longer seems as far-fetched as it does today. This would allow the customer to switch vehicle types for applications ranging from weekend ski trips, commutes to the office, and home projects involving the movement of large objects.
Most corporations aren’t ready for the transformational changes that GenAI will bring to their business. And most business leaders aren’t creating the right teams to rapidly experiment with and understand how to apply GenAI to improve productivity and generate growth. Innovation centers are already primed to effectively play these critical roles within the corporation. Innovation teams have the skills, growth mindset, and access to external-partner ecosystems to play a critical part in ensuring that corporations can stay relevant in the age of AI — as long as they follow our playbook!
Excerpted from the 2026 book The Corporate Innovator’s Playbook: Turning Ideas Into Profits, by Paul Campbell and Steffen Bartschat. (Featured image by Abbe Sublett on Unsplash.)
You must be logged in to post a comment.