Jubilant Ingrevia, an Indian specialty chemicals company, has quietly built one of the more complete examples of what AI actually looks like inside a factory. At its Bharuch site in Gujarat, home to Asia's largest acetic anhydride production line and run by fewer than 40 people, the company connected sensors, data systems and predictive software across more than 30 processes. The result: process variability dropped by 60 percent, and production volume nearly doubled. That earned the plant a place in the World Economic Forum's Global Lighthouse Network in late 2024.
The timing matters. The global chemical industry is in the middle of a prolonged downturn. Companies built more factories than the market needs, demand has stayed soft, and trade tensions have made planning harder. Profit margins that used to sit close to 6 percent on average dropped sharply starting in 2023 and have stayed low since, and most forecasters expect the pressure to continue through 2026. In that environment, squeezing more output and fewer errors out of existing plants is not a growth strategy. It is how a company protects its margins while competitors struggle.
What separates Jubilant's approach from a typical software rollout is that it did not stop at one plant or one department. The company describes its system as six connected pieces: production data and automation to boost yield, monitoring systems for pollution and emissions, AI tools to speed up product development, sensors and cameras for worker safety, training programs for staff, and predictive planning for supply chains. Each piece feeds data to the others, so a change in the supply chain shows up in production planning automatically instead of waiting for a person to notice and pass the message along.
The company has since applied the same approach to Gajraula, one of its oldest and largest facilities, running more than 25 separate plants. Over three years, the effort spanned more than 30 use cases and generated over 5 million dollars a year in value, alongside real reductions in energy use and safety incidents. A new plant, MPP8, is now being designed as a smart factory from the ground up rather than retrofitted later, which is usually far cheaper and more effective than bolting technology onto old equipment.
This is not an isolated story. The World Economic Forum's Global Lighthouse Network started in 2018 with a small group of factories and has grown to more than 200 sites across over 30 countries, spanning industries from consumer goods to steel to pharmaceuticals. The pattern across nearly all of them is the same: companies that once ran AI as isolated pilots in one department are now connecting data across the whole business, because doing it in fragments does not move the numbers enough to matter.
For any business running expensive, long-lived equipment, whether that is a steel mill, a cement plant or a chemical site, the message is blunt. A single AI pilot in one corner of the operation will not save you during a downturn. The companies pulling ahead are the ones treating data connection across departments as basic infrastructure, the same way electricity or plumbing became basic infrastructure a century ago.