Gen AI: Firms Race to Make IT Mainstream

JUST FIVE YEARS AGO, watching a sci-fi movie on artificial intelligence (AI) taking over the world or machines thinking and processing information like humans seemed too futuristic to be true. Things changed dramatically last November, when San Francisco-based OpenAI — a start-up founded by Sam Altman and Elon Musk — unleashed ChatGPT, a chatbot built using Generative Pre-trained Transformer-3.5 large language model (LLM). In less than two months, ChatGPT saw a record 100 million+ users.

Today, social media is debating whether we are closer to an AI takeover of the world. In 2021, MIT Technology Review featured GPT-3 (Generative Pre- trained Transformer-3), a LLM built by OpenAI, trained on algorithms making it capable of reading and writing like humans as one of 10 breakthrough technologies of the year. Bill Gates, founder, Microsoft, wrote in a blog that two technology demonstrations have been truly revolutionary — the introduction of graphic user interfaces like Windows and AI.

Calling ChatGPT a tech watershed is not an exaggeration. Generative AI is becoming a customer-driven value initiative that IT service companies are queuing up to crack. According to Goldman Sachs, the annual incremental global IT services gain from Gen AI will be between $123 billion and $244 billion by 2030. “Assuming a gradual adoption curve, accelerated adoption during the latter half of the decade would translate into $18-36 billion in annual global IT services spend by FY27, or 1-2% annual revenue tailwinds for our India IT services coverage by FY27,” it says. The report says Gen AI could add 50-100 bps of incremental revenue growth to Indian IT from FY25, it adds. That’s the market that IT service companies are looking to tap.

In June, Accenture was the first IT services company to call out Gen AI revenues. In the investor call, Julie Sweets, chair and CEO, Accenture, said in the past four months it had clocked $100 million in sales from 100 projects. “As we’re getting to the maturity of automation and AI before generative AI, we see generative AI as our ability to continue to give at least that 10% productivity year in and year out,” she said. Accenture plans to double its talent in data and AI practice to about 80,000. With over 300 use cases and applications across 19 verticals, Bhaskar Ghosh, chief strategy officer, Accenture, says the current client focus on Gen AI is in five areas — assisting in decisions, creating content, generation of software codes, process automation and security and protection. It has rolled out client projects in these areas.

“One of our energy clients has large data from oilfields and has to generate daily reports on performance to assess safety and security risks. It is now getting automatically generated based on the daily data using Gen AI,” says Ghosh. Accenture is using Gen AI to help Mitsui Sumitomo Insurance improve customer service experience related to accidents. This Gen AI solution works using the company’s data on insurance, laws and regulations to improve accuracy and speed in customer response.

Other companies have also dipped their toes in projects and concepts. TCS has 50 projects in concept stage and another 100 in the pipeline. One pilot in customer service assistance through Gen AI is aimed at getting accurate responses to travel insurance policy queries. “I see many opportunities for us in structuring our own way of delivery and improving value propositions that we can offer embedding Generative AI,” says Ganapathy Subramaniam, COO, TCS. It has started advising clients on creating a road map for Gen AI adoption and started training across multiple Gen AI solution suites in partnership with hyperscalers aimed at creating a Gen AI trained talent pool of one lakh employees.

In May, Infosys launched Topaz, a Gen AI-focused platform and said it was working on 80 Gen AI client projects. Gen AI is one of five key areas of focus in its bid to improve profitability. It started implementation of AI in various aspects of its software engineering processes and learning—where real-time feedback by AI on coding helps keep tabs on quality. Infosys has already trained nearly 40,000 employees on Gen AI tools to work on its open source and proprietary Gen AI platforms and modules.

Satish HC, EVP and co-head of Delivery, Infosys, says early adoption trends are being seen in engineering use cases such as code conversion, code optimisation, document extraction, summarisation, call centre and marketing content creation. “At a large investment bank, Gen AI assistants are helping wealth managers advise customers by substantially saving research time. The assistants can summarise conversations and identify actions and next steps. For a large FMCG major, we have provided personalisation by creating content for social media captions, blog articles, product descriptions and short stories,” says Satish.

How effective or accurate Gen AI will turn out to be depends on the data it is fed. The more data it has, the more it learns and the more intelligent it becomes. Here, usable data of clients becomes critical for building efficient tools. A recent McKinsey report, “The economic potential of Generative AI — The next productivity frontier,” says based on analysis of 63 use cases, an estimated $2.6 trillion to $4.4 trillion could be added annually to the global economy. If Gen AI gets fully implemented, McKinsey pegs banking, high tech and life sciences to see an annual value addition of between $200 billion and $340 billion, while retail and consumer packaged goods sectors could see $400 billion to $660 billion of value addition each year.

Subha Tatavarti, CTO, Wipro, says her company’s deep relationships with clients play a critical role in providing access to such data. Wipro, which aspires to be Gen AI-first company, is working at developing in-house platform and capabilities to drive both bottom and top line in core business processes within the company as well as in its services lines and customer experience.

“We are working with an insurance client in code generation where we are looking at deployment of 10x code developers and engineering to scale up some of their product development and reduce time to market,” says Subha. Wipro also plans to train 80% of employees in Gen AI using the DICE ID platform created by Wipro Labs to issue skill credentials through collaboration with NASSCOM and foundit (erstwhile Monster).

Meanwhile, Cognizant has over 100 clients on projects focused on cognitive and generative AI at various stages and is expected to make an investment of nearly $1 billion in generative AI capabilities over the next three years. It announced the launch of a new Gen AI focused platform, Cognizant Neuro AI, which includes a library of reusable generative AI models and tools. Through the Cognizant Google Cloud AI University programme, it aims to train 25,000 Cognizant professionals on Google Cloud AI technologies, which it also intends to extend to its clients.

HCL Tech is exploring over 140 external and internal projects in Gen AI at various stages. It has implemented a Generative AI-powered human-like voice conversation bot for a global healthcare firm that specialises in medical devices, diagnostics, nutrition products and pharmaceuticals. Even in the pre-GPT era, IT services companies were investing in building intelligent platforms and cognitive tools such as Wipro’s Holmes, Infosys’s Nia, Accenture’s AIP+, TCS’s Ignio and HCL’s DRYiCE. These platforms are the framework on which applications can be built.

Indian IT service companies are still figuring out models to create applications and tools that would be scalable. While long standing relationships with clients have given them access to required data sets in these early deals, most companies still depend on hyperscalers such as Google, Microsoft, Amazon and IBM which have built the foundational models on which Generative AI tools/application gets built.

Hyperscaler Collaborations & Risks

Microsoft, Google, IBM and Amazon have been investing in LLM on which Gen AI tools and platforms are getting built. Way back in 2019, Microsoft invested $1 billion in OpenAI to co-create and built AI technology to be leveraged exclusively by Azure and since then has increased its investment. With the latest round of reportedly $10 billion in January 2023, industry estimates put Microsoft’s cumulative investment to be around $13 billion.

Following ChatGPT’s success, Google launched its conversational chatbot Bard earlier this year, and has been working on LLMs like PaLM 2 and Gemini. IBM, whose Deep Blue supercomputer had made waves in the '90s after defeating Gary Kasparov, has also launched a Gen AI suite on its intelligence platform Watson, called WatsonX. With hyperscalers being the backbone of the IT services companies where most of their products and services are delivered, they have naturally started to deepen their collaborations to catch up and compete with their peers.

In India, Microsoft, which works with almost all services companies, is making the applications built on Azure ITeS360 by such companies Gen AI-ready. “Large use cases are also emerging in education, healthcare. That is how we are building these use cases and taking these to the market,” says Irina Ghose, managing director, Microsoft India. Google introduced Duet AI in Google Cloud and Workspace earlier this year. It aims at helping people collaborate with AI to code, write and get better insights into data. Over 7,50,000+ Workspace users already have access to the new features in a preview. “We are seeing strong demand for more than 80 models — including third-party and popular open source — in our Vertex, Search and Conversational AI platforms, with the number of customers growing more than 15X from April to June,” CEO Sundar Pichai said in the company’s investor call.

In terms of collaboration with hyperscalers, for IT service companies, it is not really about choosing one over the other. Accenture has announced a series of collaborations with all major hyperscalers. With AWS, the company plans to co-invest to develop industry-specific and cross-industry solutions, pre-built models and training for adoption of Gen AI at scale. It has also collaborated with Google Cloud. Cognizant recently said it would build a healthcare LLM solution on Google Cloud’s generative AI technology to address a range of healthcare business challenges utilising the capabilities of Google Cloud’s Vertex AI platform and its own AI domain expertise. HCL Tech has signed up with hyperscalers to offer Gen AI integrated capabilities into their HCL PromptO platform on Github Co-pilot and Google Cloud Duet AI.

The biggest concern that could hinder large-scale Gen AI adoption is security, accuracy and quality of data. The latest global study by IBM Institute for Business Value pointed out that more than half (57%) the CEOs surveyed were concerned about data security and 48% about bias or data accuracy. Manish Goyal, senior partner and global AI & analytics leader, IBM, says some internal and external risks that existed in traditional models around data laws, privacy, robustness, fairness will continue to exist even while using generative AI.

With the cry for formulating a global framework for responsible Gen AI framework gathering steam, industry hopes these concerns Would be addressed in the near future. What could be a gamechanger and challenge dominance of ChatGPT is open sourcing of these foundational LLMs which could create an unprecedented level-playing field.