What Indian managers should know about Generative AI

Given what organisations know about traditional AI models, it is time to take control of Gen AI's flaws and use its generative ability to offer complex, customised solutions

By Anjana Karumathil

4 min read

Published: Apr 8, 2024 12:51:38 PM IST

Updated: Apr 8, 2024 01:05:01 PM IST

Managers must carefully consider where AI best fits into the company’s value chain. Focus on visible gains in costs, efficiency and productivity. Image: Shutterstock

Whatever AI can do, generative AI can do better.

Generative (Gen) AI, the newest entrant to the AI toolset, has the USP of ‘generating’ outputs from user prompts in creative, unprecedented ways. This is a leap from traditional AI models that use prespecified rules to analyse and detect patterns in big data. ChatGPT4 is an example of Gen AI, and it scored 99th percentile at the US Biology Olympiad by summarising large volumes of information and answering complex questions. When ChatGPT launched in 2022, I watched in amazement as it spat out customised assignments and poetry within seconds. Within a year, we have Gen AI tools like Midjourney that create realistic images and videos from simple text prompts. However, Gen AI is not just a suite of shiny toys for enthusiasts; companies, including Microsoft, Google, and Amazon, are integrating it into their digital suites. A survey by EY predicts that by FY 2029, Gen AI may contribute over $350 billion to India’s GDP across IT services, education and healthcare.

Experts predict three emerging trends from Gen AI. First, AI models operating as intelligent cognitive interfaces with humans will increase. We already know that web searching is far more customised with ChatGPT than with Google. Bots from Amazon now recommend products and issue refunds. We are comfortable speaking into our computers, and their interfaces comprehend our words irrespective of accent. Second, Gen AI models will combine dispersed organisational data in simple, meaningful ways. Imagine you are an HR manager in an IT company. You have multiple documents on employee retention, promotion, pay and leave in different databases. If implemented correctly, Gen AI can behave as a centrepiece in your company’s technical architecture by combining relevant data in real-time. You could analyse each employee’s income, medical reimbursements, performance appraisals, and leave history in one dashboard to predict their likelihood of staying with your company and propose pay hikes. Finally, we will move past scrolling through apps with our fat fingers. Interfaces will be conversational, and talking to a computer will be as seamless as talking to a friend.

This raises the question of how managers and Gen AI must coexist in organisations. The first step is to create an AI strategy aligned with overarching business goals. Managers must carefully consider where AI best fits into the company’s value chain. Focus on visible gains in costs, efficiency and productivity. Second, managers must consult AI experts to determine the technology architecture and models best suited for the organisation. The biggest challenge is in equipping the workforce. With buzzwords like prompt engineering and RAG floating around recruitment websites, organisations may choose to hire traditional AI experts and offer targeted training to build Gen AI skills.

Also read: To use AI tools smartly, think like a strategist

Gen AI is not without pitfalls, the biggest of which may be data privacy. Entering proprietary data into generic AI models poses financial and reputational damage risks, as organisations lose control over who has access to the data and how it is deployed for training. Additionally, traditional AI models are trained using structured data customised by an organisation, while Gen AI is trained with user prompts, which, by nature, are unstructured and unpredictable. This makes Gen AI models prone to ‘hallucination’, generating erroneous responses from ambiguous inputs. Hallucinations are unacceptable in business, and managers must constantly verify training data to ensure accurate responses. Second, the training data itself may be biased. Imagine you manage a construction company that has only hired male engineers. This may have been because, historically, suitable female candidates did not apply. However, the AI model may use this pattern to interpret that women are unsuitable for this position and reject the applications of qualified female candidates who may apply in future. Finally, managers must stay informed about AI governance and ensure compliance with ever-changing regulations.

Despite these risks, it’s not just creatives who are excited about Gen AI. Given what organisations know about traditional AI models, it is time to take control of Gen AI’s flaws and use its generative ability to offer complex, customised solutions. This is time for the government to be deeply involved in building a regulatory framework for its fair and ethical use. It is also time to provide IT infrastructure to Indian startups in the AI space and build India-specific datasets to train our models. India is already the world’s IT hub; what stops us from becoming the hub of Gen AI?

Anjana Karumathil is an Associate Professor of Practice at IIM Kozhikode.