Talk to the AI without stopping to walk

The problem

What the plant lead notices on the round dies at the end of the shift.

The plant lead spends the day walking between reactors, formulation and packaging, hands busy. Nobody asks him anything because he can't type. That batch 3 shows unusual viscosity, that quality needs to be told before bottling: all of that lives in his head and disappears at shift end.

• The plant lead spends the day walking between fermenters, formulation tanks and the bottling line, with their hands busy. Nobody asks them anything because they cannot type.
• They are the one who notices that batch 3 shows unusual viscosity, that a flange on bioreactor 2 is weeping, or that a strain culture is foaming more than usual.
• All of that lives in their head. If they remember, it becomes a phone call; if they don't, it disappears at the end of the shift.
• And quality finds out about the viscosity when the batch is already in jerrycans, which is the most expensive moment to find out: the batch has to be held, sampled again and possibly reworked.

How it fits the IRIS system

Connect in voice mode — iLEAN listens to the plant lead, and answers back.

With iLEAN Connect voice mode, he dictates it without stopping: 'note that batch 3 shows high viscosity, have quality check it before bottling.' iLEAN listens and structures it as an incident linked to the active batch.

It is not a dictaphone. The plant lead says “note that batch 3 shows high viscosity, have quality check it before bottling”, and iLEAN files it as an incident on the active batch and comes back with “quality can check it at 12, shall I notify you?”. The loop closes without anyone stopping.

See the full IRIS architecture →

Before and after

The plant round today versus the round with an earpiece

Aspect	Today	With iLEAN Connect
An observation at the fermenter	In the plant lead's head	An incident on the active batch
Viscosity flagged before bottling	A call, if remembered	Routed to quality with the batch number
Reply from quality	Whenever they call back	Spoken in the earpiece, seconds later
Plant lead's effort	Stop, find a terminal	Speak while walking
End-of-shift knowledge	Gone	Queryable by batch and equipment
Conversations in the formulation hall	—	Not recorded

Seconds later, iLEAN replies with support: 'quality confirms they can check it at 12, want me to notify you?' Knowledge that used to die at shift end now enters the system in real time.

Impact estimate

Strategic value — no payback of its own, and we say so.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• No standalone payback: we do not put a number on this case because its value is not directly monetized. It is strategic, and we would rather say so than force a figure.
• Floor knowledge stops dying at the end of the shift and enters the system in real time, attached to the batch it refers to.
• Observations reach quality before bottling instead of after, which is where an unusual batch is still cheap to hold and sample again.
• And it feeds the other cases: an incident dictated on the round is one more ring of evidence when a batch is later questioned by a distributor or an auditor.

Strategic value, not directly monetized · floor knowledge stops dying at shift end.

And the fair question from the production manager

“Is this a way of recording the team?” — no. iLEAN listens to the plant lead, and only to what they choose to dictate; there is no continuous listening and no location tracking. Turning a dictated sentence into batch, symptom and recipient is an anchored task, where the best models drop below 1.5% error [1], and the plant lead hears the summary back before it is filed.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.