by Marcus Oxer, Domain Solutions Manager, Mosaic, Cenevo
While AI is driving labs toward accelerated innovation, AI adoption is requiring life sciences professionals to look at how they’re operating today. Before the true value of AI can be realized, the maelstrom of disorganized data and workflows must be tamed.
In both 2025 and 2026, Cenevo surveyed more than 110 life sciences industry professionals about AI adoption, data management, automation, and orchestration. The life sciences professionals surveyed included:
- 37% from large pharma/biotech (>5,000 employees)
- 18% from mid-size pharma/biotech (500-5,000 employees)
- 18% from small (start-up) pharma/biotech (<500 employees)
- 14% from academic institutions
- 13% from CROs (Contract Research Organizations) and others
Respondents came from R&D, discovery, chemistry, biology, clinical labs, manufacturing, industrial R&D, education, materials, and healthcare.
Of the surveyed labs, many are using generic AI tools such as ChatGPT, Claude, Google Gemini and Microsoft Copilot. Over the past year, many life sciences software vendors have introduced native AI features in their platforms, and 19% of labs are already taking advantage of these. A similar number of labs (22%) report that they are using custom or internally developed AI tools.
AI is already actively delivering insights
One of the more notable changes in the past year is the rate of adoption. In 2025, only 15% said they’d implement AI within the next year; the remainder mentioned a timeline of one to five+ years. According to the 2026 report, generative AI is already in use within more than 60% of life sciences labs, although many are still trialing it rather than it being fully deployed. Labs are turning first to AI for activities that can deliver immediate value, such as data analysis and reporting, workflow orchestration, experiment planning and sample and inventory management. Use of Agentic AI is likely to be a growth area. 27% of respondents are still in the exploratory stages of agentic AI, with 5% using agentic AI in production. We anticipate these figures will rise significantly in the coming year.
As for the future of agentic AI, scientists will see many benefits as implementation takes place. Reporting will be much less of a chore, as the agents can deliver specific formats tailored to each target audience. That will significantly streamline the ability to deliver clear information to regulators and clients (in the case of CROs/Contract Development and Manufacturing Organizations). Processing data will be simplified as natural language queries will make it easier to construct data processing pipelines. The agents can be fed scientific literature and the organization’s proprietary templates to generate specific lab protocols.
However, some labs are still lagging, with 13% stating their labs aren’t using any form of AI.
Bottlenecks remain
Lack of systems integration, for both hardware and software, is a problem in more than half of the labs. This connectivity challenge is also preventing full AI adoption.
Manual work remains a significant barrier to fully connected, AI-ready labs, but the direction of travel is encouraging. Fewer labs (33%) now rely heavily on manual processes than in last year’s survey (50%), suggesting that digitization and automation efforts are starting to deliver results. At the same time, a small but notable group of labs (14%) has moved much further ahead, with most of their workflows already automated.
Security and training need to be addressed as well. Almost 60% of scientists said they have privacy, security, trust, and compliance concerns. Slightly more than half (51%) faced skills and training challenges, with internal skills gaps among teams. In addition, certain departments may be more eager than others to adopt AI, so standardizing adoption across the organization is also a challenge.
It’s the data
Data remains one of the biggest barriers to AI adoption in the lab. Information is spread across hardware and software systems in structured, semi-structured, and unstructured formats, and quality can vary significantly. Even so, progress is being made. Fewer labs than last year cite data management, quality, and overload as major challenges, while more than half report having some form of centralized data repository or data lake in place, even if usage is not always consistent. A significant proportion (18%) say that collected lab data is stored but rarely accessed. In an AI-based future, this so called “dark data” can be ingested and used within analytics tools and made usable for both scientists and AI agents.
Agentic AI itself could help address some of the data management challenges by transforming unstructured, semi-structured, and proprietary lab data into more structured, standardized formats that are easier to analyze.
Follow the money
In 2025, 65% of respondents said that they would be spending more on inventory management software. With the accelerated pace of AI adoption, budget priorities have changed, instead targeting the infrastructure necessary to streamline adoption. Automation platforms, such as robotics, autonomous mobile robots, and liquid handlers are the top spending priorities. AI-enabled software is a close second, followed by systems integration, which focuses on connectivity among instruments, electronic lab notebooks, and lab information management systems. Data analytics and infrastructure remain important, but their lower position in this year’s spending priorities may reflect the fact that many labs have already applied AI to data analysis and are now shifting attention toward the automation, integration, and operational foundations needed to extend AI into other areas.
Digital lab maturity
The share of organizations describing themselves as fully digitalized has fallen from 16% to 9%, surely reflecting a shift to a new perspective as to what a truly digital lab represents.
However, the digitization varies among the types of respondents: 21% of large pharma/biotech organizations say they’re fully digitized while none of the mid-sized biotechs or smaller consider themselves to have fully digitized environments.
Connectivity, automation and AI
The connected lab is moving closer to reality. As labs move beyond standalone software tools, the next phase of progress will depend on combining automation, orchestration, AI, and data analytics to accelerate research.
Agentic AI could support this shift by simplifying complex operational tasks. For example, automation agents could turn workflows into deployable automation configurations without specialist programming or API skills, while inventory agents could help scientists locate materials across lab and organizational collections, reducing duplication and supply costs.
However, significant practical barriers remain, including challenges of instrument integration, incorporating legacy systems or in-house software, and addressing regulatory requirements.
Over the next year, labs expect to apply AI across practical areas such as experiment design and planning, sample and inventory management, software development and scripting, and reporting or documentation.
Large pharma organizations appear to be targeting operational transformations, with particular focus on areas such as workflow automation, assay optimization, and instrument integration.
The future of AI in life sciences
AI adoption has moved from hype to reality. The focus has initially been on more easily achievable realities of data analysis and workflow orchestration ahead of longer-term transformational plans.
Now that organizations are starting to implement AI, they’re addressing the steps necessary to streamline existing lab practices. Both budgets and time are being devoted to making organizations more AI “ready,” including investment in everything from automation platforms and instrument integration to building data lakes and replacing legacy systems.
The physical-virtual environments within life sciences labs make AI adoption a unique challenge that many other industries don’t face. By streamlining existing workflows and processes in the “real world,” labs will be able to adopt AI more effectively, both for their physical and software-based discovery.
The survey report can be downloaded here.
About the author
Marcus Oxer is Domain Solutions Manager at Cenevo, which delivers lab management systems, automation, orchestration, data management and AI technology for life sciences.