For me, data is like instruction. If you lead a team and give misguided instructions, you can imagine how the team will perform. Data is the same.
Shailja TewariChief Data Officer and VP CSQ, Schneider ElectricAI can help industrial companies predict problems, like a part that’s about to fail, before they reach the customer. But a prediction is only as good as the data behind it. In most industrial companies, data is scattered and inconsistent across manufacturing, service, and sales systems that were never built to share information. When AI pulls from those disconnected sources, its answers can sound right and still be wrong. Useful AI starts with connecting that data and making someone responsible for its quality.
Shailja Tewari is Vice President of Customer Satisfaction and Quality and Chief Data Officer at Schneider Electric India, the local arm of the global energy-management and industrial-automation company. Her dual role puts her in charge of both the company’s data and the customer-facing quality outcomes it drives. That’s why she treats data integrity as a first-order engineering problem.
“For me, data is like instruction. If you lead a team and give misguided instructions, you can imagine how the team will perform. Data is the same,” Tewari says. At Schneider, that means connecting data across manufacturing, quality, service, and sales, and building a trusted source of truth before relying on AI to make predictions.
Tracing failures end to end
In an industrial company, the full story behind a problem rarely sits in one system. Manufacturing has the component and process history. Service knows what happened during installation or commissioning. Quality records the failure and its root cause, and sales and customer teams hold the rest. Because each team only sees its own piece, the shape of the data often limits AI more than the model does. “What’s sitting with manufacturing versus services versus sales is very different, and it’s not talking to each other,” Tewari says.
Her team pulls those pieces onto a single platform so a problem at a customer site can be traced back to manufacturing, engineering, or the supply chain. “Individually, these data points provide information. Connected, they provide intelligence,” she notes. With the data connected, the team can go beyond asking what went wrong. They can ask where the same problem might happen next and how to prevent it. That’s the move from reporting on the past to predicting the future.
The data also has to be accurate, and Tewari is clear about where hallucination comes from. “With low data quality, the output will be directionless. The output will have a lot of hallucinations,” she says. When people log what happened on a job, they need to record it carefully. If the record doesn’t match what someone did on site, every recommendation built on it will be off. Good data won’t stop hallucinations completely, but strong data quality and governance make AI results far more accurate and reliable.
Data quality gets owners
Everyone agrees data quality matters, which makes it easy for no one to own it. A central data office can’t carry that responsibility alone. The teams that create and use the data have to own its quality too, so Tewari has made it someone’s job in every part of the business. “Every function, every process has one data business partner who takes care of the data quality,” she notes.
The approach is federated, with accountability sitting close to where the data originates. Finance owns its financial data, sales owns its customer and commercial data, and manufacturing and quality own their operational data. Each team knows its own information better than a central group could. Tewari says the approach is helping “by far,” and the company recognizes the people who produce clean data so good habits get noticed.
Governing the source of truth
A central Data Office ties those teams together into one enterprise data system. “We worked with third parties who gave us a framework and direction, and our internal team has been instrumental in setting up the framework, the architecture, how it looks and operates, and where the data lies,” Tewari adds.
Reliable data is what makes people trust the output. A support engineer using AI to diagnose a problem needs to know the recommendation is based on validated information. A quality leader looking at a predictive signal needs to know where the data came from and who is responsible for it. Without that foundation, AI stays an experiment and never becomes part of everyday work. The stronger the data layer underneath, the more useful the AI built on top of it.
When manufacturing, service, quality, and sales data come together, teams can spot patterns that no single function would see on its own. That helps them catch quality issues earlier and find root causes faster. “The ultimate value of connected data is the ability to anticipate an issue before the customer experiences it. When data helps us move from detection to prediction and from prediction to prevention, that is when AI begins to create real value for quality and customer trust,” Tewari says.





