Imagine a smart voice assistant — like Alexa, but for the factory floor — that guides workers step by step through complex manufacturing tasks. When a new employee starts, instead of months of expensive training, this assistant coaches them in real time and catches quality problems before they become costly rejects. It even explains its own reasoning so workers actually trust and understand the AI suggestions. Three real factories (textiles, home appliances, packaging) already tested it.
Manufacturing companies lose significant money on product defects and spend months training new workers, especially as experienced staff retire. The skilled labor shortage — driven by demographic change — hits hardest in knowledge-intensive production where quality depends on human expertise. Companies need a way to preserve institutional knowledge, get new hires productive faster, and catch quality issues before they reach the customer.
A privacy-focused digital voice assistant built on the open-source Mycroft platform, with prescriptive quality analytics and a WHY engine for explainable AI. Two versions of the manufacturing demonstrator were delivered, plus the standalone WHY engine prototype — all tested across three industrial use cases.
If you are a textile manufacturer dealing with an aging workforce and rising defect rates — this project developed an AI voice assistant that guides workers through complex production steps in real time. Tested directly in an Italian textile factory, it targets a 30-60% reduction in failure costs. The system explains its recommendations so operators trust and adopt it, instead of fighting it.
If you are an appliance producer struggling with quality control and long training cycles for new hires — this project built a digital assistant with prescriptive quality analytics that catches problems before they become expensive rejects. The assistant shortens worker training time with an expected reduction in changeover time of 15-30%. It was validated in a real white goods production environment.
If you are a packaging company losing production time every time you switch between product runs — this project developed an AI-assisted training system that gets operators up to speed faster on new configurations. Tested in a liquid packaging facility, the system targets 15-30% reduction in changeover time. The voice-first interface means workers keep their hands free during operation.
The COALA assistant is built on Mycroft, an open-source voice platform, which reduces licensing costs compared to proprietary solutions. Specific pricing for commercial deployment is not published in the project data. Contact the consortium to discuss implementation costs for your specific production environment.
The project delivered two versions of the manufacturing demonstrator, with version 2 described as ready for full integration and first exploitation activities. Three different industrial environments (textile, white goods, liquid packaging) validated the system, suggesting it handles diverse manufacturing contexts. Scaling to full production would likely require customization to your specific processes.
The consortium of 17 partners across 5 countries developed the technology under an EU Research and Innovation Action. IP ownership typically follows the Horizon 2020 grant agreement, meaning each partner owns the IP they generated. Contact the coordinator BIBA in Bremen, Germany, to discuss licensing or partnership options.
The project specifically chose Mycroft as its base because it is a privacy-focused open-source assistant — unlike commercial alternatives that send voice data to external cloud servers. AI ethics was a core design principle addressed during design, deployment, and use of the solution. The project also developed a didactic concept to educate workers about opportunities, challenges, and risks in human-AI collaboration.
Based on available project data, the system went through two development iterations over the three-year project. The project includes a concurrent change management process designed to support adoption. Expect an implementation period that includes customization to your production processes, worker onboarding, and the change management activities.
The digital assistant core was designed for integration into manufacturing environments, with version 2 explicitly built for full integration readiness. The voice-first approach means minimal hardware disruption — workers interact through speech rather than new screens or interfaces. Specific integration requirements would depend on your existing IT and production infrastructure.
The project ended in September 2023 and involved Digital Innovation Hubs to replicate demonstrators across Europe. The consortium includes 10 industry partners and 6 SMEs, suggesting commercial motivation beyond pure research. Contact BIBA to learn about current support options and which partners are pursuing commercial offerings.
The COALA consortium is unusually industry-heavy for an EU research project: 10 out of 17 partners come from industry (59%), including 6 SMEs, which signals strong commercial intent rather than purely academic interest. The remaining partners — 4 universities and 3 research organizations — provide the scientific backbone. Spread across Germany, Greece, France, Italy, and the Netherlands, the consortium covers major European manufacturing markets. The coordinator BIBA in Bremen is a well-established production and logistics research institute with deep industry ties, making them a credible bridge between research results and factory adoption.
BIBA - Bremer Institut fuer Produktion und Logistik GmbH, Bremen, Germany. Use SciTransfer's coordinator lookup service to get the right contact person.
Want to connect with the COALA team about licensing or piloting their AI manufacturing assistant? SciTransfer can arrange a direct introduction and help you evaluate fit for your production environment.
