How L&T is engineering an AI-driven conglomerate

This is just one example of how India’s largest engineering and infrastructure conglomerate has been using digital technologies such as radio frequency identification (RFID) tags, geofencing, internet of things (IoT) devices, drones, robotic process automation, and even augmented reality (AR) and virtual reality (VR) tools to keep a hawk eye on its assets across the country.

These tools are all aimed at reducing waste and pilferage while increasing productivity and growth of L&T’s low-margin businesses.

“Digitalization is certainly impacting our business, qualitatively and quantitatively, tangibly, and intangibly,” said S.N. Subrahmanyan, who became the group’s chairman and managing director on 1 October 2023 after A.M. Naik stepped down from his non-executive chairman role.

Digitalization “is empowering teams to predict better, improve productivity, reduce costs and wastage, make more informed decisions, enhance efficiencies, and deliver faster to customers,” Subrahmanyan said in an interview with Mint.

People, materials and machines

L&T’s digital journey broadly revolves around materials, machines, and people.

For one, L&T’s command centre, which provides a one-stop dashboard for Subrahmanyan when doing reviews of the group’s leaders, is connected to about 14,000 devices (of a total of 30,000) on an IoT platform. This dashboard also enables team leaders to detect fuel pilferage at any of the company’s 600 sites within 10 seconds, since all the fuel that is dispensed gets tracked and captured with the help of active RFID tags. 

L&T can also digitally measure the exact quantity of concrete that gets produced at its plants with the help of sensors, giving its business heads a complete footprint of the concrete produced, transported and poured.

Also read: From hiring engineers to India’s best technology, L&T boss S N Subrahmanyan speaks his heart out

The conglomerate also has more than 300 fully automated digital weighbridge systems and computer vision cameras to measure the net weight of the ₹1.2-billion worth of materials that it purchases annually, which includes concrete or cement, steel, diesel, aluminum and copper. 

The group has also onboarded 2.5 million contractual workers into its so-called Worker Induction and Skill Assessment (WISA) system, which helps managers track the location of each worker, gauge their daily  productivity, and even suggest training to upgrade their skills.

Millions in savings

By leveraging the strengths of its group companies including LTI Mindtree, “who are our system integrators”, and L&T Technology Services (LTTS), “which focuses on product engineering”, the savings from the use of these digital technologies is huge, said Mahesh Chikodi, the group’s chief digital officer. 

“Just 1% savings in fuel today amounts to nearly $9 million; 1% saving in concrete saves us roughly around $12 million; and 1% savings amounts to about $19 million in steel that is the costliest.”

Also, given that L&T tracks all the hardware and assets such as cranes, earth-moving equipment, cameras, and road rollers it has “deployed even in the remotest parts of India”, “we get to know if any pilferage has happened within seconds. We then get a snapshot and send it to the hiring agency (contractors) and penalise them”, said Chikodi. 

According to him, L&T recovers around ₹4 crore from hiring agencies in terms of penalties every quarter. “Other than getting money back, it also dissuades more pilferage,” adds Chikodi.

Embracing AI, GenAI

L&T aspires to become an artificial intelligence-driven company to further its digital journey. 

“The ultimate destination of our journey of digital transformation is to significantly enhance the efficiency of our expansive operations, widely implementing a suite of digital solutions, conceived, and developed in-house, based on artificial intelligence, machine learning, IoT, cloud computing, predictive analysis, big data, geospatial technologies, and more,” said Subrahmanyan. 

He cited risk-prediction as a case in point, “where AI is already making a huge difference in our ability to assess, pre-empt and combat risk”.

L&T has “moved from a stage of descriptive analytics to predictive analytics, and now to prescriptive analytics”, added Chikodi. 

The group has already been using machine learning in areas of supply chain, finance, human resources, customer relationship management, and project management. It can, for instance, predict the failure of a machine part “at least three to four weeks in advance because of the data that we collect”, and alert the business heads concerned and equipment vendors to take preventive action.

L&T has also begun exploring the use of generative AI tools. For instance, team leaders have begun using Microsoft’s AI-powered tool called Copilot for meetings and presentations. 

“Whenever any meeting takes place in L&T, we use Copilot to write the minutes (instead of someone doing it manually) before the meeting ends. Hence, the participants know who has to take what actions. We will also be using Copilot in Power BI (business intelligence) reports. Historically, what people used to see is just descriptive dashboards, but using Copilot you can ask queries (in natural language) and it will give you reports,” explained Chikodi.

A super app for L&T

L&T has an internal name for its AI journey—Larsen and Toubro Cognitive Services (LNTCS). According to Chikodi, L&T’s AI journey will be completed in three phases.

The first was to fine-tune large language models (LLMs) for contracts and legal documents, which is “now live”. The next step is to use AI and ML for supply chain, finance, HR, CRM, and project management, for which L&T is using a platform from LTIMindtree called Canvas AI. 

For instance, the HR department uses AI and ML to measure the performance levels of all employees, carve a career path for them, and explore what training they would need.

“The last bit is the most complex, where we will use diffusion models for design automation. This will happen in the next few months,” said Chikodi. 

When asked to build a bridge, for instance, L&T engineers would look at existing Autodesk drawings to identify the most complex project and the learning from those projects, which would take months. 

Diffusion models, on the other hand, are helping L&T convert PDF documents into drawings. And all such data—be it from enterprise resource planning (ERP), information technology, operations technology, logistics data, metadata, or from third-party applications—converge into a single data warehouse, and can be accessed through a “super app”. 

“Any slicing and dicing of data will happen through our LNTCS platform,” said Chikodi. This implies that L&T’s business leaders can ask any question in, say, English, such as: “Give me the profitability of a specific project from its inception date till now” to get a response on the dashboard.

To ensure that its digital momentum is sustainable, L&T has also set up a “Digital Council” comprising a little more than 50 leaders from across group companies, who meet every quarter and discuss their priorities for the next six months to a year.

Leveraging humungous amounts of good data

All of L&T’s AI and GenAI initiatives leverage the strength of good data. According to Chikodi, L&T has gathered nearly 900 TB (terabytes) of structured and unstructured data from its machines (sensors and cameras included) that currently reside on a public cloud, Microsoft Azure. (A single TB can hold 250 movies or 500 hours of HD video.)

L&T plans to move this data to its own two data centers—one in Kanchipuram near Chennai and the other in Panvel, Mumbai. 

L&T is also building digital twins, or simulations of physical assets, by scanning the geographies where it has projects with the help of helicopters and drones. All the geo-data is hosted online to help design engineers and clients to simply log in, view the geography in the form of satellite imagery or a 3D model, and reuse the data as required for the group’s projects. 

“The same data from a geography that was scanned for a specific independent company, be it a water or power transmission distribution unit, can be reused for another purpose,” explained Chikodi.

These digital initiatives have allowed L&T to not just improve process control but also its ability to adopt emerging technologies, said Subrahmanyan. “The L&T that you see is in reality wired very differently with new-age technologies embedded across functions for efficiency, excellence, and growth.”