Artificial intelligence (AI) is transforming research, development, and innovation (R&D&I), unlocking new possibilities to address some of the world’s most pressing challenges, including sustainability, healthcare, climate change, and food and energy security, as well as help Organizations enable organizations to help organizations to help organizations challenge better innovation and launch breakthrough products and services.
Artificial intelligence in R&D is nothing new. However, the rise of generative AI (Genai) and large language models (LLM) has greatly expanded its capabilities, accelerating breakthroughs and overall innovation.
How can organizations benefit from AI in R&D and I efforts? What are the best practices to take on success? To identify Arthur D. Little’s (ADL)’s Blueshift Academy conducted a comprehensive study, interviewed more than 40 AI providers, experts and practitioners, and conducted surveys of more than 200 organizations in the public and private sectors. . Results report, Eureka! About Steroids: AI-driven research, development, and innovationin-depth analysis of the current landscape and future trajectory of AI in research and innovation.
Our analysis focuses on five key areas:
Artificial intelligence offers benefits in R&I – but it won’t replace humans
Every building block of R&D&I can benefit from AI, from technology and market intelligence to innovation strategies, conceptions, portfolio and project management, and IP management. When we want to understand these benefits, three key factors emerge:
- Artificial intelligence will increase researchers, rather than replace them, free up time and make them more productive and creative
- Artificial intelligence helps solve difficult problems that cannot be tried due to the speed and expansion and learning ability of technology, opening up new avenues of innovation
- AI will take on the role of “planner thinker”, moving beyond content generation and search to cover more complex roles such as becoming a knowledge manager, hypothesis generator, and assistant to the R&D&I team.
There is no blanket model to deploy when deciding whether to use AI to address specific R&D and usage situations. To understand which AI approach will make organizations focus on two factors – the type and amount of available data (from slightly to a lot) and the nature of the questions to be asked (from open to specific). Meanwhile, a single AI approach may not provide the best results – most of the most advanced intelligent systems produced in the past 15 years are systems of systems. These are independent AI systems, models or algorithms designed for specific tasks, and after being merged, they provide greater functionality and performance.
Success requires eight good practices
Based on interviews with researchers, AI scientists, founders and persons in charge of R&D for digital, manufacturing, marketing and R&D teams, we see eight good practices for successful AI deployments. Organizations need:
- Adopt an agile approach so that teams can work quickly in a rapidly changing AI environment
- Build a strong foundation by focusing on data quality, collaboration of organizations, and leveraging proprietary data
- Make strategic choices between building, buying and fine-tuning models, and the latter approach is often the most effective
- Consider analytical tradeoffs to ensure progress in proof of concept projects such as acquisition and synthesis of data, precision and recall, and underfitting and overfitting
- Proactively utilize available data science talents, including working with organizations outside of their organization to acquire the necessary skills
- Keep in line with it to balance safety and compliance with safety speeds
- Show benefits quickly and get user buys to build trust and unlock further investments
- Continuous maintenance and monitoring of system performance, especially around model improvements
3. Technical components are now in place
Like most AI use cases, R&D and I value chains include three layers – infrastructure, model developers, and applications.
In terms of infrastructure, it is expensive to implement and maintain sufficient computing power, but hosting providers are increasingly providing inference models to run inferences and queries in the cloud to eliminate the need for internal infrastructure and reduce internal Infrastructure needs upfront costs and democratization to obtain AI.
The value chain of AI in R&D&I relies heavily on the main open source models of players such as Meta, Microsoft and Nvidia. But smaller players like Mistral and Cohere, like academic institutions, form a key part of the ecosystem.
On the application side of the chain, general and professional R&D&I applications have been created to meet most use cases, and now there are over 500 available, covering the entire R&D&I Process.
The future is not clear – but the plan helps understand
R&D & how I will develop AI depends on the outcomes of three main factors – performance, trust, and affordability. Combining these factors, six reasonable future solutions on the spectrum between the R&D and I aspects are formed between AI, and are used only in selective, low-risk use cases. From the maximum impact to the minimum impact, these are:
- Blockbuster: Throughout the R&D cycle, AI became the primary one, reshaping the organization along the way. Data becomes the new boundary.
- Pleasant person: AI is convenient, affordable, and used for everyday productivity tasks, but fails to provide science/create value.
- Crown Jewelry: AI offers breakthroughs in productivity and science, but only targets those affordable organizations – leading to the two-speed world of R&D&I.
- Question child: Despite some iconic use cases and affordable solutions, AI has failed to prove its value – R&D and I organizations remain focused on data security, morality and lack of interpretability.
- Best Secret: AI performance has improved, but high costs have made organizations more risk averse. Low trust and traditional Chinese tape sections are restricted to adoption, and few new bold experiments are initiated.
- Cheap and annoying: AI is widely used in low-risk use cases, but is only used as a prototype or brainstorming tool. The system of distrust is strictly reviewed and the output is verified, thus reducing productivity gains.
Understanding these solutions is important for R&D and I organizations to move forward for AI adoption.
The time for R&D and I organizations to take action is
In some cases, AI can already increase double-digit time, cost and efficiency in formulation, product development, intelligence, and other R&D&I tasks. This means that in either case, six reorganization-free actions will help R&D and I organizations build resilience and leverage the benefits of AI. They need:
- Manage and empower talent to ensure workforce training and expertise on AI, if necessary
- Control what AI generates, update risk management processes, and publicly share verification methods to build trust
- Build data sharing and collaboration to work with the wider public and private sector ecosystems to drive successful adoption of AI
- Long-term training to educate the broadest user base, required skills and potential risks on AI fundamentals
- Rethinking organization and governance, taking it beyond it to provide advanced concerns and breaking down silos for smooth collaboration
- Interactive resources, working with partners or sharing resources internally to meet current and future infrastructure needs
In addition to these non-restructuring moves, success will be successful, from creating a balanced portfolio of AI-based R&D and I investments, aligning with the company’s goals. This means considering the scope, costs and benefits of a specific AI use case and using this scope to drive optimization of the innovative portfolio. Decisions should be based on strategic goals, capabilities and market intelligence and the context of organizational operations.
Every stage of the value chain of research, development and innovation can be transformed through artificial intelligence, thereby enhancing human researchers to transform productivity and achieve breakthrough innovation. These opportunities need to be balanced with a range of challenges surrounding performance, trust, and affordability, meaning organizations must immediately focus on positioning their R&I AI efforts in order to succeed, no matter what comes in the future.
This article was written with the help of Albert Meige, Zoe Huczok, Arnaud Siraudin and Arthur D. Little.
