Four Pillars for AI in Research-Intensive Work

Research-intensive industries such as pharmaceuticals, aerospace, and finance have increasingly adopted artificial intelligence (AI). This raises questions about the scientific and engineering method, accountability, human judgment, and organisational transformation. Yet, previous scholarship has addressed these topics in isolation, across separate disciplines.

At CLAIR 2026 we propose a framework that treats these four topics as interdependent pillars. The first pillar concerns AI and the Scientific and Engineering Method. It addresses the epistemic layer and asks what counts as discovery, validation, and explanation (Krenn et al., 2022; Messeri & Crockett, 2024). This question becomes acute when generative models propose drug candidates faster than scientists can account for why those candidates work (Ren et al., 2025).

The second pillar concerns Leadership, Governance, and Accountability. It addresses the institutional layer and asks who answers, to whom, and for what (Bovens, 2007; Mittelstadt et al., 2016). Aerospace developers are now expected to demonstrate accountability for model behaviour ahead of safety-relevant deployment (European Union Aviation Safety Agency [EASA], 2024; European Parliament & Council, 2024).

The third pillar concerns Critical Thinking in Algorithmic Environments. It addresses the cognitive layer and asks how human judgment is preserved when reasoning is shared with machines (Mosier & Skitka, 1996; Parasuraman & Riley, 1997; Lee et al., 2025). Recent evidence from clinical radiology shows that practitioners accept confident AI outputs even when the underlying imaging contradicts them (Dratsch et al., 2023).

The fourth pillar concerns Organisational and Cultural Transformation. It addresses the organisational layer and asks how roles, workflows, and incentives co-evolve with AI (Trist & Bamforth, 1951; Faraj et al., 2018). Moderna's enterprise rollout of generative assistants in pharmaceutical development has already prompted governance redesign around approval authority for custom-built AI tools and their use in clinical decisions (Bojinov et al., 2025).

Each pillar fails when isolated. Methodological rigour without governance produces reproducibility theatre. Governance without cognition becomes checkbox accountability. Cognition without organisational redesign rewards offloading. Organisational change without epistemic discipline transforms culture around a hollow core.

CLAIR's 2026 contribution is integrative: a cross-industry, leadership-oriented forum that brings the four topics into the same room to form the joint analysis the field currently lacks.