Innovation project

Voice-to-Action: the last mile of shop floor digitalisation

Operators work hands-busy and eyes-busy. Voice-to-Action turns what they say on the floor into a validated action in the workflow, without sending anyone to a terminal.

Book a Demo See the Platform

Operator crouching beside packaging equipment with both hands occupied during a clean, inspect, lubricate routine

Digitalisation on the floor stops at the operator.

Backend systems keep getting smarter. The physical step between the operator and those systems does not. Registration still means walking to a terminal, unlocking a screen, and typing while the line runs.

So registration happens late, or not at all. Context disappears with it: which machine, which moment, which work order. What one shift solved, the next shift solves again because the answer stayed in a notebook.

Voice-to-Action closes that last step. Speech becomes the input, the system does the paperwork.

What changes on the floor

The innovation is not speech recognition on its own. It is the chain from a spoken observation to a traceable action, without losing the context in between.

Today Observation and registration are separated by time

Observation The operator sees the problem
Decision The operator knows what is needed
30 to 120 minutes, or a shift change Time between detection and registration
Registration At a terminal, later or never
Action Maintenance, follow-up, correction

Observation and registration are disconnected
Context is lost: machine, moment, work order
The action arrives late, or cannot be traced back

With Voice-to-Action Observation, decision and registration collapse into one step

Spoken on the floor Captured on the edge, noise filtered out
Enriched with context Machine, work order, role, shift
Translated to intent A structured proposal, not raw text
Validated in the Voice Inbox The operator confirms with a single tap
Action Issue, task or notification, with audit trail

Information latency drops from more than an hour to minutes
Context is captured at the moment it happens
Every action traces back to who said what, and when

What the project builds

Reliable capture

Industrial wearables and edge processing that isolate a voice command in environments with high background noise, and keep working when connectivity drops.

Context-aware recognition

Transcription that uses live operational context, work orders, machines and plant jargon, instead of treating every sentence as an isolated piece of audio.

Intent engine

A layer that reads meaning, not just words, and turns it into a concrete proposal inside an existing shop floor workflow.

Voice Inbox

A low friction validation step. The operator stays the editor in chief, confirms or corrects, and every confirmation writes a verifiable audit trail for ISO and HACCP.

A cross-border partnership

Neither company could build this alone. The hardware and audio expertise sits in the Netherlands, the workflow and adoption expertise in Flanders.

Maecos Belgium, lead partner: The execution and orchestration layer. Context, intent, the Voice Inbox, and the integration into existing shop floor processes.
Taggl Netherlands: Industrial edge hardware and audio processing. Wearables, noise suppression and a transcription pipeline built for noisy production environments.
ERVO Projects Belgium, practice partner: Defines the industrial use cases and runs the pilot in a 24/7 metalworking environment with a highly digitalised machine park.
A large food producer Belgium, second validation site: Tests the solution in a consumer packaged goods context, with high volumes and strict quality and compliance requirements.

Experienced operator explaining a valve assembly to a younger colleague on the production floor

What we expect to prove

Speech that survives the shop floor. Accurate transcription under high noise, with dialect, accents and personal protective equipment in the way.
Intent that holds up in production. Recognition that stays reliable across real industrial use cases, not only in a lab.
Latency measured in minutes. The time between an observation and a registered, actionable item drops from more than an hour to minutes.
Adoption without training overhead. Speech removes the digital threshold for operators with a practical background or a language barrier.
Knowledge that stays. Tacit know-how becomes reusable in context, so it survives shift changes and retirements.

Results are validated in two production environments during the pilot phase. We will publish what we learn here as the project progresses.

Project details

Project: Voice-to-Action (V2A), the last mile in shop floor digitalisation through voice intelligence
Programme: Interreg VI Flanders-Netherlands, CrossRoads Vlaanderen-Nederland scheme, theme Industry 4.0
Partners: Maecos (BE, lead partner) and Taggl B.V. (NL), with ERVO Projects and a large food producer as practice partners
Duration: 15 June 2026 to 14 June 2028

This project is funded by the European Union through the Interreg VI Flanders-Netherlands programme (European Regional Development Fund), under the CrossRoads Vlaanderen-Nederland scheme for cross-border innovation between small and medium sized companies.

Lees deze pagina in het Nederlands

Curious how this lands in your operation?

Voice-to-Action is being built on top of the Maecos platform. Book a demo and we will show you what runs today, and where this is going.

Book a Demo → See the Platform