AI-Ready Starts with Data-Ready: Highlights from LabTech Europe 2026

‍

At LabTech Europe 2026, Ben Greeves of Uncountable spoke about what pharma organisations need in place before AI can deliver real value. His main point: many companies call themselves AI-ready, but few are data-ready, and that gap decides who moves beyond pilot projects.

The next leap in software: Software has moved from command line to desktop, cloud and mobile. Large language models are the next step, and they change how people work with software. Users describe goals to the software instead of telling it what to do step by step. In the lab, that means asking for "the analytical experiments I ran yesterday" instead of building a filter by hand. It also means asking for the relationship between yield and reaction time instead of configuring a chart.

Lab data, though, is still spread across every earlier era: paper batch records, desktop spreadsheets, cloud ELNs and data lakes, and limited mobile capture. AI can't reason across information it can't see.

‍

What changes for scientists

‍As routine work is automated, scientists move from carrying out workflows to directing them. Ben showed two examples in Uncountable:

• Structuring data automatically: Bodie, Uncountable's AI assistant, turns a free-text notebook entry for a 100 mg tablet into a structured experiment record. The record lists each ingredient, quantity and process parameter, and the scientist reviews and approves it.
• Generating reports: from a single prompt, Bodie analyses 21 experiments, finds a negative correlation between reaction time and yield, and writes a report with a chart. Every conclusion links back to the source experiments.

‍

Three pitfalls to avoid

1. Messy or siloed data. AI amplifies inconsistency. If one cell line appears under several spellings, an LLM can't reason reliably across it.
2. Security and IP risk. Use proper access controls and audit logs. Too much caution, though, slows adoption.
3. Disconnected integrations. AI must be built into existing lab processes, not added on separately.

‍

How to prepare

• Adopt new tools faster. Retire legacy tools, not just add new ones.
• Give everyone access. Put LLM tools in the hands of scientists and technicians, not just data teams.
• Centralise and standardise. Connect R&D and quality data so AI sees the full picture.