Historically, AI has advanced in fits and starts. But today we are experiencing a technological inflection point with a surge in AI-derived applications for R&D.
This was certainly the sentiment expressed during McKinsey’s inaugural North America Research and Development Leaders Forum hosted in San Jose, California, in May. CTOs of several companies took to the stage, sparking vigorous conversation among 40 industry leaders on topics such as AI-driven innovation; accelerating product development; governance; and operating model selection.
Four key takeaways emerged:
1. R&D innovation needs a holistic AI strategy to succeed
Many attendees agreed that owning the technology strategy is crucial, and it should prioritize use cases with a clear road map that aligns with core strengths of the company. One effective technique is creating a lighthouse use case—a compelling application that energizes and inspires your workforce to generate momentum. To get started, build early solutions in a “sandbox” environment and pilot; this has the benefits of speed and risk mitigation before scaling the technology across many aspects of the organization.
Critical questions the panelists highlighted for strategy definition include:
- What are the most impactful use cases for our business?
- How do we measure ROI and impact for these use cases?
- What technology requirements exist to achieve impact at scale (such as infrastructure, latency, capacity)?
- Who are the most critical stakeholders to define each element of our AI strategy?
2. Both analytical AI and gen AI offer compelling paths to R&D impact
A consensus from the event is that R&D leaders do not see AI as a lever to cut R&D budgets, but as a tool to supercharge outcomes and accelerate the pace of innovation. According to our survey of R&D leaders, expected impacts include:
- A 20 to 50 percent increase in product‒market fit (for example, by integrating AI to expand and explore the solution space of design concepts)
- A 15 to 60 percent increase in product performance (for example, by leveraging deep-learning surrogates to evaluate performance of various designs more quickly)
- A 30 to 50 percent increase in workplace productivity (for example, by using a virtual R&D expert to identify root causes faster)
- A 20 to 40 percent reduction in time to market (for example, by implementing gen AI to accelerate coding and computer-aided design generation)
The use cases surrounding simulations and deep-learning surrogates created a lot of excitement. AI surrogate models are thousands of times faster than traditional physics-based simulations (such as computational fluid dynamics and finite element analysis), giving users the ability to test a much larger set of potential designs. Generative models can automatically create and iterate designs and go beyond traditional parameterized approaches.
Some applications shared included a CPG company doing material selection ~70x faster and an F1 racing team modeling air flow in their design process with simulation speeds ~10,000x faster than traditional methods.
3. With AI becoming more integrated into digital engineering workflows, a day in the life of an engineer will soon look very different
AI is enabling impact across phases of the product development process, from idea generation to production. A managing director from a software company shared how implementing generative AI in the engineering development process led to a 48 percent reduction in validation efforts and a 55 percent reduction in design time.
Another transformative example shared was where AI was used to ingest the corpus of RFPs and initial designs to shorten the response time, leading to a 30 percent increase in win rate. Such acceleration of product development means that engineers will get to a functionally superior output faster than in the past and can explore a broader solution space (exhibit).
4. One of the biggest challenges to implementing AI is governance
This topic created significant debate among the panelists. While some believed a top-down implementation is the route to success, others felt business functions need to drive their own pace of implementation because they are at different levels of tech maturity and capability. Ultimately, a mix of both is sure to be required, and close integration of all teams—from physical engineers to digital and AI experts—is needed. Building AI can’t happen with a bunch of smart people “off in a corner” of an organization; that is a recipe for stifled innovation with little or no organizational momentum.
Governance that enables accountability while coordinating across business units and different functional areas is one of the biggest challenges to scaling AI and digital engineering.
Inspiring growth and efficiency
For bold R&D leaders diving into AI, a holistic strategy, robust operating model, and governance approach should be the aim. Company-wide transformations are the most effective way to fully realize AI’s potential. They help achieve alignment on high-impact use cases, develop an understanding of the required technical infrastructure, and establish change management workflows across business units. To gain momentum, nothing wins like success. Start with quicks wins and snowball them into a cascade that moves the transformation journey along.
A final message from the forum? It is imperative for R&D leaders to act swiftly in adopting AI strategies to remain competitive and capitalize on emerging opportunities in their industries.
This is a critical moment where we will be left behind if we don’t develop AI capabilities in engineering.
The authors wish to thank Alex Gopinathan, Ben Meigs, Benjamin Braverman, and Richard Bates for their contributions to this blog post.
Be part of the conversation with McKinsey Talks Operations. Learn more and register for events with our Operations experts here.