The Strategic Bottleneck: How Generative AI is Reshaping Corporate Innovation

Paper by Almog Goldstein – July 2026 – View Publication on SSRN

Founding Executive Director, UC Berkeley Open Innovation Labs

I would like to contribute this paper to UC Berkeley Professor Sara Beckman, a Product Management & Design Thinking legend, who taught and keeps teaching me how to observe the world, how to search for the truth instead of validating what I think is right, how to frame observations into insights, and remain humble by questioning my own assumptions.

Paper Summary

This paper, based on insights from over 50 innovation and R&D executives, highlights three critical trends shaping the generative AI landscape:

  • FOMO-Driven, Strategy-less Adoption: Companies often rush to implement AI without clear strategic objectives, leading to inefficiencies and a focus on “being AI-driven” rather than solving core business problems.
  • The Bottleneck Shift: Rapid R&D acceleration has shifted the organizational bottleneck from execution to upstream ideation, strategic direction, and rigorous review of AI-generated work.
  • Competitive Landscape Fragmentation: Lower barriers to entry allow AI-native startups to disrupt traditional incumbents. Large firms must improve their “sense” capabilities to track fragmented market signals and respond with agility.

Background & Purpose

Since the beginning of 2026, I’ve had meetings with over 50 innovation and R&D executives working for leading global companies across various industries. Out of curiosity, I’ve asked questions to understand how they feel about the generative AI revolution. At some point, patterns started to emerge. I’ve heard the same answers in different ways, and I was able to label these observations into three recurring themes.

The purpose of this paper is to share my observations with both academics and practitioners from the industry, so perhaps it could be a building block for new initiatives, strategies, or thinking frames that would create more abundance that does good for humanity.

I’ve used Claude Sonnet 5 for the process of analyzing and summarizing my observations.

Labeled Themes

Across the meetings, participant observations cluster into three recurring themes.

1. FOMO-driven, strategy-less AI adoption

The most repeated observation: companies are applying AI “to everything” – R&D, Sales, Customer Support, Operations, IT, Logistics, etc. out of fear of missing out, without a clear problem definition or strategic rationale – often at a cost exceeding the human alternative it replaces. Obviously, no well-established company wants to be eaten alive by an AI-native firm. But, AI is a product. We “hire” a product to complete a job or a set of jobs. It could be a functional, social, or emotional job. When internal adoption of AI for the sake of being AI-driven becomes the main strategy, we make our employees work for AI, instead of having AI work for them to accomplish our mission.

  • During a public keynote, where I presented these observations, an attendee from an industrial company said her default approach is to “feed AI into every process” as a first pass, testing after the fact whether it was the right call. A second attendee pushed back, warning that without staying focused on business value, “we are wasting too much time playing with the different tools.”
  • Another executive warned that “accelerating bad processes just produces faster bad processes” and that efficiency-driven AI adoption without strategy overlooks KPIs, financial readiness, and organizational capability.
  • Two more executives both cited AI adoption as outpacing internal governance and R&D-strategy definition.

2. The bottleneck shift: from engineering execution to ideation and review

I remember working as a Product Manager for a software startup, knowing when I leave for a vacation I can leave 2-3 well-defined features, backed by detailed PRDs and UI/UX designs, and upon my return the R&D team will still be working on them. Today, startups producing mobile apps are able to move faster than ever – but do they run in the right direction? Product Managers can’t just come up with random tickets to fill the R&D’s backlog – time is money. How can we make sure that the initiatives given by Product Managers to the R&D teams are aligned with the corporate strategy and long-term goals?

In addition, engineers have a much higher stake in the product decision-making process. While you can have full control over every single detail in the platform, today with a single prompt on a leading Generative AI tool you can have a fully designed platform and fine-tune it. I would recommend giving more UX and commercial context to your engineers so they can make better product decisions. They are an integral part of the product development process, and treating them as such could benefit the company.

A second, closely related observation raised in nearly every meeting: AI has made R&D/engineering execution so fast that the constraint on companies has moved upstream – to ideation, strategic direction, and reviewing AI output – rather than building things. R&D used to be the bottleneck. Now in some organizations, the weight is moving from R&D back to the reasoning process – back to asking strategic questions. Product Managers “used to be able to build a couple of features and know R&D would need a month or two,” but now R&D outpaces the strategy that is supposed to direct it.

  • As one executive has mentioned – engineers have shifted from being judges of innovation initiatives to becoming their primary owners – a structural role reversal the company attributes directly to AI integration.
  • A CTO for a startup that has been acquired by a major firm had described his role as becoming “primarily a reviewer of AI-generated code,” with AI-accelerated R&D degrading output quality and institutional knowledge.
  • Another executive has mentioned that AI tools are making it easier to generate output but harder to ensure quality – developers produce code they cannot explain,” with CTO-level roles shifting toward constant review work.

3. Competitive landscape fragmentation

A third recurring observation: AI has lowered the barrier to shipping products so far that tracking the competitive landscape – who the new market players even are – has become difficult in itself. Large organizations risk being disrupted by small startups that can release full products in months (or sometimes, even weeks), as executives at large corporations are still internally realigning their structures and processes.

Leveraging Professor David Teece’s Dynamic Capabilities theory, and specifically the Sense capability, means that firms need to automate their signals collection to be able to track the market and validate weak signals when appropriate to frame opportunities.

  •   An executive cited as a cautionary tale of moving too slowly – Tesla and BYD captured the EV market while large firms watched, and now non-traditional entrants are seizing EV-charging infrastructure ahead of traditional oil and gas companies.
  •     During our June 2026 Open Innovation Advisory Board meeting, it was noted, by contrast, that AI-native startups (e.g., OpenAI) are “naturally dynamic from inception,” making small bets and iterating in ways that large firms structurally struggle to replicate.

Quotes from participants

1. FOMO-driven, strategy-less AI adoption

  • “We know that AI will help us to do it much better, more efficiently, but we must still be focused on the value and the results that we have to obtain. If not, we are wasting too much time playing with the different tools”
  • “Obviously there’s a desire or need to start using AI, but then it’s the how, it’s the what, it’s the why. Are we going to see value in this, or are we just going to slap it onto something so we can say it’s AI-enabled? Even some things, to be honest, could just be solved with basic automation, but then we use AI because it’s what’s expected.”
  • “It’s just such a buzzword, and people just do it because of FOMO, as you mentioned — because they don’t want to be left behind. They just want to do it because everybody is doing it. But there is a strategy missing on how to use it, and why to use it, and when to use it.”
  • “It got shuffled. The space became messy — and specifically, for me, the mess comes from the enormous amount of fakes, whether on social media or elsewhere.”

2. The bottleneck shift: from R&D execution to ideation and review

  • “R&D is catching up to where marketing is on a product-development side. But I believe ideas are the easy part. The real tough part is getting the question right — what’s the question that we’re trying to ask?”
  • “AI is great but what it’s going to do is degrade the quality of all of our stuff, and we’re not going to have as good of technologies out there. We’re being asked to go quicker, and we are — but what it’s doing is not giving the time for the fundamental, big, extensive technology development, because science can’t be rushed. AI is great, but it doesn’t accelerate technology development like we may want it to.”
  • “If you use technology to accelerate a bad process, it’s going to be a fast, bad process. That doesn’t help… It is important to be efficient, but it’s not a strategy.”
  • “The very easy takeout for me is that engineering became the commodity, and that’s it. If you accept that fact, you need to find the competitive advantage somewhere else.”

3. Competitive landscape fragmentation

  • “If you look into all of the management literature, you’ll find that most companies don’t survive because they become very efficient, and their efficiency makes them myopic — and then they’re easily disrupted.”
  • “It’s going to be very disruptive. There’s going to be a lot of creation, but a whole lot of disruption or destruction — the creative destruction is going to be different than it was before.”

I would love to hear your thoughts on this paper! Please feel free to reach out at [email protected].

About the Open Innovation Labs: The Open Innovation Labs at the UC Berkeley Haas School of Business serve as a hub for research, education, and industry collaboration, dedicated to advancing the theory and practice of open innovation.

How to cite this work: > Goldstein, Almog, The Strategic Bottleneck: How Generative AI is Reshaping Corporate Innovation (July 12, 2026). Available at SSRN: https://ssrn.com/abstract=7105378 or http://dx.doi.org/10.2139/ssrn.7105378