Our convictions for R&D and Innovation in 2026
Functions across the enterprise are facing the same dual pressures right now: Artificial Intelligence (AI) is rapidly changing how work gets done, and geopolitical uncertainty is making the operating environment less stable. Research & Development and Innovation (R&D+I) is no exception.
Now that we’re a few months into 2026, we’re starting to see how these pressures are translating into R&D+I’s day-to-day work: teams are learning how to integrate AI into core workflows, inputs are becoming less predictable, and more is being asked upfront, with less room for iteration later.
Given this trajectory, we thought it’s time to set out our key convictions for R&D+I in 2026.
Not as predictions, but as a reflection of what’s already taking hold, and what’s likely to matter more as the year unfolds.
Reach out to discuss this topic in depth.
Conviction 1: Simulation and autonomous experimentation will displace a significant share of early testing: R&D+I will shift more front-end work into digital and automated environments to accelerate iteration and reduce physical validation cycles
Simulation and autonomous experimentation are not entirely new. For years, industries such as pharmaceuticals and automotive have used simulation models and automated testing environments to explore design options and reduce the need for physical trials. However, what is new is how early and how heavily teams are beginning to rely on these technologies.
Work that used to be de-risked through physical testing (such as formulation screening, parameter tuning, or performance trade-offs) is increasingly being pushed upstream into simulation and automated experimentation. These approaches now act as a first filter, allowing teams to narrow options digitally before committing to physical validation.
Primary drivers include compressed development cycles, reduced tolerance for downstream rework, and increased expectations for decision accuracy earlier in the development process.
However, the downside risks of this transition are real. Model quality, data integrity, and result interpretation carry greater weight much earlier in the lifecycle. Without the right guardrails in place, organizations risk locking in flawed assumptions before they are fully tested or understood. As a result, moving forward, greater emphasis will be placed on how models are built, validated, and maintained, in addition to iteration speed.
Conviction 2: Geopolitics and sustainability will become hard constraints in early input decisions: Upstream decisions will increasingly be shaped by supply stability, regulatory fit, and sustainability criteria from the start
These factors have always been part of R&D+I, but their impact is showing up much earlier in the process and becoming harder to work around.
For instance, trade restrictions and regionalization are fragmenting supply in ways that are difficult to ignore. The situation with critical minerals and rare earth elements is a clear example, forcing companies to rethink sourcing strategies as access varies significantly by region. At the same time, sustainability requirements are ruling out certain materials and inputs earlier in the process. Many companies are phasing out hard-to-recycle plastics such as multi-layer packaging, forcing teams to redesign products around materials that meet recyclability standards.
As a result, input choices are increasingly treated as constraints from the outset, meaning the “best” option is not always available or allowed. Decisions around materials, ingredients, suppliers, and so on are being shaped less by what is ideal and more by what is consistently accessible and compliant across regions, limiting the solution space before development even begins.
For R&D+I, the shift has been subtle but increasingly hard to ignore. What used to be optimized later (for example, through reformulation or sourcing changes) now affects what gets developed in the first place.
Optionality is beginning to matter more, and substitution is being baked in, leaving fewer assumptions about what can be fixed downstream.
Conviction 3: R&D+I roles will evolve into hybrid profiles combining scientific depth with AI fluency: the “AI-augmented scientist” will become more common, as expectations grow for teams to interpret model outputs and work alongside intelligent tools as part of everyday practice
This is not about AI replacing domain expertise; it’s more about changing how that expertise is applied.
As in other functions, AI is becoming embedded in R&D+I’s day-to-day work, from analysis to design support and documentation. As a result, teams are increasingly expected not only to engage with model outputs but also to understand their assumptions and limitations.
The bar for R&D+I roles is shifting from optional AI use to required fluency, with clear implications for how talent is sourced and developed: greater emphasis on upskilling existing teams, integrating AI literacy into core training, and redefining role profiles to reflect hybrid expectations. The differentiator is no longer just technical depth, but the ability to apply judgment and experience and, importantly, knowing when to trust and when to challenge the model.
Conviction 4: Institutional knowledge will become a live input for daily decisions. Large volumes of past experiments, decisions, and failures often remain buried across electronic lab notebooks (ELNs) and local systems. But, AI models and agentic systems that surface relevant history in real time will start to shape how teams assess feasibility and frame new ideas.
Most R&D+I organizations are sitting on years of prior work (such as experiment outcomes, decisions made, and approaches that failed) but that information has historically been difficult to access. It can be fragmented across systems and too time-consuming to retrieve and interpret in context.
However, the ability to query and surface that knowledge in a usable way is changing. Advances in AI, combined with improved data integration, are making it possible to bring relevant past work into the flow of decision-making, reducing reliance on manual search or individual memory.
As a result, more of that history is surfacing when teams scope new efforts or decide whether to move forward. Instead of starting from scratch, they are working against what has already been tried.
In some cases, it is as simple as realizing “we have already done this” and choosing not to spend the next six months repeating it. The net effect is a shift away from rediscovering the past and toward work that actually moves things forward.
Conviction 5: AI governance will be embedded into R&D+I to ensure controlled, transparent, and responsible model use: Guardrails around data quality, model performance, and appropriate use will become non-negotiable as AI integrates into core development workflows
As AI becomes embedded in R&D+I work, teams will need a clearer understanding of where model data comes from and how models are trained to build confidence in the reliability of their outputs.
In practice, this means putting stronger guardrails in place: ensuring data quality, monitoring model performance, maintaining traceability of outputs, and defining clear boundaries for when AI should and should not be used.
Without these guardrails, decisions risk being made without a clear link back to the underlying data or logic, making them harder to explain, challenge, or improve. Thus, in 2026, governance will move to the forefront of R&D+I priorities, influencing how teams adopt and rely on AI.
As we head into Q2
As we head into Q2, we’ll see these convictions show up in very tangible ways.
AI will be even more embedded in day-to-day workflows. Geopolitical pressure will continue to influence design decisions. More effort will shift upstream, and downstream flexibility will continue to decrease.
The cost of getting decisions wrong won’t just be technical; it will increasingly be strategic. What that means in practice is that advantage will no longer be just about speed; it will be about maximizing decision quality earlier in the lifecycle.
And that’s a very different capability than most R&D+I organizations were designed for…though they’re adapting quickly.
If you enjoyed this blog, check out our, R&D and Innovation Overview – Everest Group Research Portal, which delves deeper into other topics relating to R&D and Innovation.
If you’d like to continue the discussion about the topics featured in this blog in more depth, please contact Jillian Walker ([email protected])