[Adobe Stock]
In practice, however, most labs are still dealing with manual workflows, fragmented systems, and siloed data. The reality is that AI can’t deliver its promised value without automation and strong data foundations.
This article explores why automation and data management are prerequisites for AI, how automation alone isn’t the answer, and what labs need to address today to prepare for tomorrow.
AI relies on reproducible, high-quality data
AI is only as good as the environment it operates in. In the lab, the lack of integration between lab software and hardware means data tends to be collected in silos with different formats. For AI to reliably produce insights across those data sources and not just within them, data should be machine-readable and interoperable. Applying FAIR (findable, accessible, interoperable, and reusable) principles to data is as important as ever when considering AI1.
Data needs to be traceable and well-structured, covering the entire process lifecycle. This allows results to be unambiguously associated with specific samples. This may include information such as where a sample originated, how it was processed, what instruments it was processed on, and at what time. Timestamp capture is critical because it enables other data sources (e.g., experimental metadata and environmental conditions) to be considered within the analysis alongside sample and experimental data, potentially generating deeper insights.
For maximum benefit, AI needs access to data that spans analytical instruments, laboratory information management systems (LIMS), electronic lab notebooks (ELNs), sample inventory systems and automation platforms. This requires a level of integration among software platforms or, at least, shared currencies in the form of sample identifiers, container barcodes, and timestamps. It also demands a level of lab orchestration whereby software-based experimental design flows seamlessly into automated execution in work cells, and results are captured directly by connected systems – thus avoiding manual data handoffs that can introduce human error.
Automated workflows excel at generating large volumes of consistent, high-quality, timestamped data and eliminate human variability. Stable and repeatable automated processes produce clean data. When manual workarounds creep in so does noisy and unreliable data.
Automation comes with its own challenges
By design, automation enforces structure and machine-readability. While automation is fundamental to unlocking the potential of a digitalized lab, the reality is more nuanced. The accuracy and repeatability of automation are key to success.
In drug discovery, where processes are often highly miniaturized, any calibration errors or deviations can have significant impacts, such as pipetting failures, which may only be picked up when the data is analyzed or, worse, lead to the wrong conclusions and misdirected follow-on experiments.
Automated laboratory systems, from liquid handlers to complex workflow cells, require regular maintenance, calibration, and monitoring. If errors start to appear, the impact can ripple quickly, resulting in downstream decisions based on flawed data. Furthermore, the higher throughput nature of automated processes amplifies the consequences of errors if they are not detected early. Companies such as Recursion Pharmaceuticals have explored novel ways to monitor their automation in real time using robotic vision to identify errors early on and mitigate this problem3.
Fortunately, good data management and AI help here. Connecting timestamped automation usage data to experimental outcomes can help identify patterns that highlight potential issues developing within the automation workflow. The key is to identify problems early enough to avoid misleading trends and wasted experiments. A loss of trust in automation, potentially causing teams to revert to manual rework, can hinder efforts to speed lab operations.
Siloed data limits AI’s oversight
Organizations frequently have islands of automation and software platforms across laboratories or departments, managed and maintained independently. AI thrives on correlations across systems. Without harmonization between or across these systems, AI only sees fragments of the full picture and misses the relationships that drive meaningful insights.
Therefore, AI may behave inconsistently or give incomplete results, failing to scale and reach its potential. Automated systems depend on metadata consistency: plate layouts, sample identifiers, timestamps, batch information, etc. If these fundamentals are not standardized and managed, even the most advanced automation will generate data that AI cannot reliably interpret.
Put simply, if samples, plates, lots, or workflows are identified differently across platforms or systems, AI cannot map outcomes to origins. For this reason, good data governance, metadata standards, and controlled vocabularies are now as essential as automation. Free-text fields, varied naming conventions, and spreadsheet-based workflows do not have a place in this vision.
Building an AI-ready lab: tips for success
1. Strong data governance: Consolidating fragmented IT systems and inconsistent data formats is often the biggest hurdle, yet it is fundamental to success. Standard identifiers, structured metadata, and clear data ownership give AI the high-integrity inputs it needs.
2. Move foundational systems to cloud or SaaS: Cloud-based laboratory platforms make upgrades faster, improve reliability, reduce total cost of ownership, and offer modern APIs that streamline integration with AI tools. Cloud platforms benefit from continuous updates and robust security, allowing teams to concentrate on strategic objectives instead of routine infrastructure upkeep.
3. Understand the strengths and limitations of automation: Ensure measures are in place to account for potential automation errors.
4. Don’t forget routine maintenance and validation: Capturing machine utilization data helps with maintenance planning; many platforms have this built in. Results may be affected when equipment isn’t performing optimally.
5. Invest in change management: Automation and AI require shifts in mindset, not just technology. Empowering advocates, communicating the “why,” building trust, and delivering early wins are keys to success.
AI success needs strong foundations
Laboratories do not need to be perfect to start working with AI, but they do need to be digitally mature enough that automation and data are reliable, and systems are harmonized.
Marcus Oxer
Automation and AI without strong foundations may amplify any weaknesses. When data foundations are solid, automation coupled with AI can accelerate discovery in ways previously impossible.
The labs that invest in building automation on the foundations of good data governance and harmonized digital workflows will be the ones that realize the true promise of AI tomorrow.
Marcus Oxer is Domain Solutions Manager at Cenevo, which specializes in lab management systems, automation, orchestration, data management and AI technology for life sciences.
References
- National Institutes of Health (NIH) / National Institute of Allergy and Infectious Diseases (NIAID) (n.d.) FAIR Data Principles at NIAID. Available at: https://www.niaid.nih.gov/research/fair-data-principles (Accessed: 10-Feb-2026).
- An, Z., Xu, B., Bruske, J., Dai, X., Han, C. and Wang, S. (2024) A survey of offline precision calibration of industrial robots. Advances in Computational Engineering, 5(2). Available at: https://ace.ewapub.com/article/view/16572 (Accessed: 19 February 2026).
- Recursion Pharmaceuticals (2024). Improving Wet Lab Automation with Computer Vision. Available at: https://www.recursion.com/news/improving-wet-lab-automation-with-computer-vision-a-hack-week-breakthrough (Accessed: 10-Feb-2026)
