Researcher at glovebox

Source: © CasarsaGuru/Getty Images

Half of business leaders surveyed said they had made important decision before all the data was in

Almost two-thirds of business R&D leaders have made important R&D decisions using flawed data, while around half have made decisions without all the evidence, according to a new survey. The majority of respondents expect AI will transform R&D decision-making over the next five years and predict that combining AI with trusted evidence will give organisations the edge over those that don’t embrace AI.

Elsevier asked more than 1000 R&D leaders, managers and researchers at R&D-led corporations how they use data and evidence to make decisions and how AI is being applied. The report found that just over three quarters (76%) of respondents are confident in the quality of their data. Yet 63% say their organisation has at some point made an important R&D decision using data or evidence later found to be inaccurate, incomplete or outdated. Just over half say their organisation has had to make an important R&D decision before all the evidence was in.

The survey also shows that the use of AI is not yet routine in R&D decision-making. Under half say it is used regularly, while a further 40% say it is being piloted. Almost half of respondents use general-purpose AI tools such as ChatGPT or Copilot for literature searches, against 23% who use specialist scientific AI tools. Specialist tools are the most common choice for data analysis and modelling.

Most respondents (81%) expect AI to fundamentally transform how R&D decisions are made over the next five years while 86% believe organisations that combine AI with high-quality evidence will outperform those that do not. However, just under half say implementing AI at scale will be a significant challenge.

‘Confidence is not the same as accuracy,’ says Stuart Whayman, president of corporate markets at Elsevier. ‘R&D teams are under pressure to decide faster and often before all the evidence is in. Confidence keeps them moving, but it can’t make up for evidence that is incomplete, outdated or unreliable. AI will raise the stakes, because it will amplify whatever evidence it is given, good or bad. The advantage will go to the organisations that can bring their evidence together and trace and verify what sits behind their decisions.’

Leaders in the pharma/biotech sector are more confident in data quality (85%) than the average (76%). Similar numbers say their organisation has made important decisions on evidence later found to be flawed and before all the evidence is in. However, more people in pharma/biotech use AI for scientific writing, reporting or documentation (79%) than in other sectors (68%) and for scientific modelling, simulation or experimental design (74% compared with 66%).

‘I’m not surprised that managers make R&D decisions before all the evidence is available,’ says Philip Shapira, an innovation specialist at the University of Manchester, UK. ‘Uncertainty is inherent in research, particularly in its early stages, when neither the likelihood of success nor the eventual value of the work is clear. Waiting for complete evidence could mean never acting. The challenge is to seek out reliable information, recognise what remains unknown and adapt as new evidence emerges.’

Shapira believes generative AI can be helpful to managers, but its outputs always need to be treated cautiously and critically. ‘It does seem likely that generative AI will be increasingly used to aid R&D managers to find and synthesise information and explore possibilities. But it will be unlikely to replace human judgement, creative spark and willingness to take risks that successful R&D requires.’