Back in May 2024, Tata Consultancy Services (TCS) carried out a study which revealed businesses wanted to use AI for innovation and revenue growth, but were not sure how to do it.
Now, TCS has teamed up with Nvidia to help accelerate AI adoption across a range of industries by using the latter’s infrastructure to build customized AI solutions for sectors including manufacturing, automotive, finance, and retail based on large language models.
I spoke with Anupam Singhal, President of Manufacturing at TCS, about how these AI tools work, how they can benefit businesses, and how to mitigate the risks of using AI (if any) in these industries.
When you think of language models, the topic of manufacturing or automotive industries doesn’t jump straight to the front of the mind. Can you tell me more about the link between language models and how they can help manufacturing and a range of other industries?
Whilst language models might initially seem more suited for textual tasks and we do often corelate B2C use cases to them, they also possess significant potential to revolutionise industries like manufacturing and automotive in B2B and B2B2C cases.
We are already seeing this potential come to life through our Future Ready Manufacturing solutions. Here, we are transforming repair and service cycles and predictive maintenance too. We use language models to analyse historical data and identify patterns which can help indicate potential equipment failures, reducing downtime and optimising maintenance schedules. What’s more, using Generative AI and SLM’s (Specialised language Models) we can transform the daily activity of a Repair and Service Technicians to improve the time taken in the repair/service cycle.
Language models are also transforming supply chain resilience. By analysing supply chain data these models can optimise inventory levels, improve logistics, and mitigate supply chain disruptions.
TCS finetunes LLMs based on industry expertise. What goes into this process and what safeguards are in place to ensure that the model does not provide sub optimal suggestions for industry challenges
Finetuning LLMs for industry-specific applications is a meticulous process. Firstly, using our industry expertise and customers’ ecosystem, we curate a high-quality dataset specific to the industry, ensuring it covers a wide range of scenarios and edge cases. We then leverage pre-trained and out-of-the-box language models as a foundation, and fine-tune these based on the industry-specific data sets.
Our engineers then play a crucial role in monitoring the model’s performance, providing feedback and making necessary adjustments. By feeding in intuitive adjustments and tacit knowledge this uplifts the accuracy of the model. Once this insight has been gathered and fed into the model, we then implement strict ethical guidelines to ensure that the model generates fair and unbiased outputs. Appropriate guardrails are defined with TCS’s agent-based monitoring approach to ensure the responses are continuously monitored.
Our recently launched TCS Manufacturing AI for industrials suite, built on NVIDIA’s tech stack, is the perfect example of a fine-tuned LLM/SLM for this sector.
Whilst our technology is advanced, we see it as a means to a destination, not a destination itself. Its core purpose is to transform the daily lives of citizens. Our solutions are designed to assist humans, not replace them. Whilst AI can handle routine tasks, analyse vast amounts of data, and identify patterns that may be otherwise missed, human judgment, creativity, and critical thinking remain essential for making complex decisions and providing strategic direction.
Our goal is to create a synergistic relationship where AI complements human capabilities, enabling us to achieve greater efficiency and innovation. For example, we helped a large building materials firm put a quote in the market 1.5x faster by using generative AI to provide critical insights on the right configuration and automating the draft quote creation. This bolstered the productivity of those working – rather than replacing them.
While it’s true that LLMs can produce outputs that mimic training data and, in some cases, data generation can be synthetic, LLMs are also able to generate novel insights and creative solutions for a variety of different industry problems. By analysing vast amounts of data and identifying underlying patterns, these models can uncover hidden opportunities and innovative approaches that may not be apparent to human experts.
For example, TCS is currently building a solution called Mobility AI, part of our TCS Mobility Suite. One use case of this solution is deciphering a parking sign in microsecond intervals and recommending the next best action. These insights enable decisions that humans cannot make at that micro-moment in the journey of the driver, highlighting how GenAI can transform this experience in the future of Mobility.
However, it’s important to note that AI is a tool, and its effectiveness depends on how it’s used. Human guidance and oversight are crucial to ensure that AI is used responsibly and ethically.
Do we run the risk of losing human expertise if we rely more heavily on AI?
Far from diminishing human expertise, AI can enhance it. By automating routine tasks, AI frees up human workers to focus on higher-value activities that require creativity, problem-solving, and empathy. Moreover, AI can provide valuable insights and support, enabling humans to make more informed decisions. The key is to embrace AI as a tool that empowers humans, rather than a threat. By working together, humans and AI can achieve remarkable things.
