Manufacturing IT leaders now can leverage technological advances that enable Microsoft AI assistants to become a human-friendly interface to shopfloor systems. Machine operators, mechanics, technicians, and safety inspectors can talk to a specialized GenAI Microsoft copilot in natural and native language. The copilot responds – via voice or text – in the same language, with answers to questions and recommendations for action. The copilot can intelligently source the recommendations from a combination of multiple documents such as SOPs, FMEAs, RCAs, and troubleshooting manuals.
The benefits are immediate: IT leaders don’t have to build shopfloor AI capabilities from scratch. Instead, they can evaluate specialized AI plugins from systems integrators such as Tata Consultancy Services (TCS) in partnership with Microsoft AI. These plugins can be tuned and deployed in weeks instead of months.
The right data at the right time with targeted AI
A common problem for shopfloor workers is quickly finding information to help with a machine problem or process. Information typically lives in different locations and databases, in different formats, including physical or online manuals that run to hundreds of pages. Much of this information is in English, which is a challenge for workers who are non-native English speakers.
The time spent finding – and understanding – answers to questions is time that manufacturing machines lie idle or run inefficiently.
But properly trained and adapted Microsoft AI copilots can help plant workers break through this information blockage. Using a smartphone or simple online voice or text query, workers can ask natural language questions. The copilot answers specifically and contextually and can recommend actions to take. AI-trained translation services present these results in the workers’ native language and can even intelligently mix native and English terms when appropriate.
To achieve this, TCS conducts a “Gemba walk” (from the Japanese word for “real world”) with a customer’s operators and workers to understand what they do and how they do it. The operators’ actions are used to configure the copilot’s underlying large language model (LLM) and the natural language prompts workers will use.
This data integration can potentially be complex – but it’s a key step in making AI effective. To produce specific, accurate, contextual responses, the copilot needs access to various data sources, both structured (ie., conventional RDBMSs) and unstructured (ie., manuals and diagrams). Data may need to be replicated from isolated sources into a data lake. Partnering with an experienced integrator like TCS can speed up this entire process when combined with Microsoft Cloud, Azure and AI capabilities.
What to expect from the AI “co-worker”
AI has its own behaviors and both IT staff and shopfloor workers should know what to expect when interacting with a copilot. Examples:
- AI models can “drift” – meaning the AI response slowly degrades over time, often when real-world manufacturing data differs from training data.
- Models can also “hallucinate” – as in, produce a response that is completely unexpected. It might be an output that is nonsensical, offensive, inaccurate or non-existent.
These are known problems. A systems integrator such as TCS can work with IT staff to monitor and remediate these problems. TCS can also work with shopfloor workers, using proven techniques, to help them evaluate the copilot’s response.
The bottom line
Picking the right AI copilot partner gives manufacturing IT leaders the capability to make work plant easier, faster, and more efficient. In fact, TCS has seen Microsoft copilots help cut mean-time-to-repair (MTTR) by 10-15% monthly. Such savings can directly increase the manufacture’s bottom line by up to 2%.
Manufacturers need to rapidly leverage the power of AI and GenAI for competitive advantage. System integrators like TCS can handhold them through the learning curve and develop solutions based on Microsoft AI to extract business benefits, combining the power of strong domain understanding and the latest technologies.
