Faster Innovation with AI in R&D Starts with Your Data

Where AI already pays off in R&D, and why data readiness and governance decide who captures the gains.

AI is changing research and development (R&D) quickly, and the clearest example sits inside the companies that build AI. By August 2026, Anthropic reported AI systems completing 26% of its internal AI R&D work with only high-level human supervision, up from 1% five months earlier. The figure appears in a September 2026 report published by the Cambridge Programme on AI Science & Policy (CASP) at the University of Cambridge. Its co-authors include Geoffrey Hinton, Yoshua Bengio, and senior scientists from OpenAI, Microsoft, and Anthropic.

Business leaders are already planning for the shift. In an IBM Institute for Business Value study of 2,000 C-suite executives, 79% expect AI to contribute significantly to revenue by 2030. Nearly half of AI spend goes on efficiency today, and respondents expect 62% of it to fund innovation by 2030. The prize is smarter, faster innovation, with new technology as the means. Here is what that looks like in R&D, and what decides who captures it.

What an intelligence explosion means

The Cambridge report asks what happens when AI automates the work of building AI. Could progress accelerate into an "intelligence explosion", where years of advances are compressed into months or less? The authors find preliminary evidence that it could. Today, only thousands of researchers work on frontier AI, the most capable systems. Because AI systems can be copied and run in parallel, automating that work could add the equivalent of millions more.

The report is written for policymakers, but one observation matters for business. The authors argue that AI R&D may well be automated before most other work, because it is a largely digital domain with clear measures of success. In our view, much of corporate R&D shares those traits, from simulation and software to requirements engineering, which makes it a natural place for AI to deliver early.

Where AI already pays off in R&D

R&D teams can act on three opportunities now.

• Faster innovation cycles. AI speeds up experimentation, variant management, and requirements engineering. In its February 2025 guide to AI-powered R&D, Boston Consulting Group (BCG) expects AI adoption in R&D to cut time to market by 10-20%.
• Lower costs. BCG expects the same adoption to bring R&D costs down by up to 20%. It sees the full effect once companies rework their operating model to match.
• Agentic workflows. AI agents are software that plans and carries out multi-step tasks towards a goal. CASP reports that the best systems now complete AI R&D tasks that take human experts hours to days, up from seconds-long tasks in 2023. Agents that take on that kind of background work free R&D teams to focus on strategy, judgement, and creative growth.

Data readiness is a leadership decision

Every one of those gains runs on data. Gartner predicts that through 2026, organisations will abandon 60% of AI projects unsupported by AI-ready data, meaning data that represents the use case and is governed and documented well enough to trust. In its survey of 1,203 data management leaders, 63% of organisations either lacked the right data management practices for AI or were unsure whether they had them.

That makes data readiness a leadership issue. Which data an R&D programme needs, who owns it, and how it is governed are business choices first and technical tasks second. BCG lists a new data architecture, with dedicated AI and data layers, as one of five main levers for leading in AI-powered R&D. A strategy that names the use cases, the data behind them, and the target key performance indicators (KPIs) gives that data work a clear brief. We cover the groundwork in why AI agents need the right data foundation.

How data and governance gaps stall agents

The hard part is scaling. RAND cites estimates that more than 80% of AI projects fail, twice the rate of IT projects without AI. Its interviews with 65 data scientists and engineers point to misunderstood problems, missing data, and weak data infrastructure among the leading causes. We map those traps in how to make your AI project pay back.

Agents raise the stakes because they act as well as answer. BCG's 2026 research found that only 5% of companies have the full set of critical controls in place, and companies with all six controls generate three times the agentic AI value of those with one. We looked at what that means for design in our take on BCG's findings.

The Cambridge report gives an example from the frontier. Roughly 1,200 OpenAI agents, tasked with cyber evaluations in isolation, coordinated over a makeshift message board, obtained unauthorised internet access, and hacked into Hugging Face. Clear permissions, audit trails, and human checkpoints are what let agents take on more work with confidence. The same discipline covers security, compliance, and the cost of software and compute, and it keeps the return measurable from the first pilot.

The bottom line

AI helps R&D teams run more experiments, analyse data faster, and bring products to market sooner. Integrated well, it reduces costs, increases speed to market, removes unnecessary complexity, and opens new sources of competitive advantage. The organisations that capture those gains treat data quality, data availability, security, and software costs as the core of the programme from day one. Part two of this series covers talent and the workforce; regulation, security, and ethics; costs and return on investment (ROI); and personalisation.

The Cambridge researchers expect much more automation of AI R&D over the next few years, and the foundations you set this year decide how much of it your R&D can use. If you are planning where AI fits in your R&D roadmap, we can help you choose the first workflow, test whether your data is ready, and design the controls that let agents scale. Book a call and we'll start with the workflow that matters most to your business.

Sources. CASP, What if automating AI R&D triggers an intelligence explosion?. IBM Institute for Business Value, Enterprise 2030. BCG, AI-powered R&D. Gartner, Lack of AI-ready data puts AI projects at risk. RAND, The root causes of failure for AI projects. BCG, The formula for agentic AI value.