The Gap Between Collecting Process Data and Using It

In this session

1. 2:48Jack on why agentic AI is a different kind of tool: it plans and executes, it does not autocomplete.
2. 14:30The six kinds of data a bioprocess engineer needs to see, and the system each one usually lives in.
3. 24:00Scores for 32 sites and 50 products, and what a green, yellow or red plant feels like to work in.
4. 33:30Six barriers, from can we measure it to do we act on it, and where agents move each one.
5. 37:48Masaki on trust: inspect the data, the method and the evidence, then validate the workflow for its context of use.
6. 44:30Harmonised data, a harness around the model, domain agents: how Invert is built, and a root cause exploration triggered by a titer forecast.
7. 58:44Audience Q&A: AI for process control, whether a model can be validated, what to ask an AI vendor, LLM business continuity.

About this session

Jack Prior, Head of Process Monitoring and Data/AI Strategy at Sanofi, on measuring how ready a manufacturing network's process data really is: a maturity model scored across 32 sites and 50 products, why maturity gets harder as networks digitise, and the six barriers between a sensor reading and a decision. Masaki Yamada, Head of Product at Invert, on what it takes to put that data to work: trust you can inspect, a harness around the model, and agents that recommend while a human decides. Then the questions the audience asked them both. Recorded September 16, 2026 with BioPharma Webinars.