Boosting productivity is only the sideshow for AI: Transforming good work into great will be its real benefit

2023 will be remembered as the year generative AI became mainstream.

Nothing demonstrates the technology’s impact on our public consciousness better than leading dictionary Collins naming “AI” as their word of the year. The need to regulate AI was hotly debated, along with its benefits and impact on the economy and jobs.

The World Economic Forum’s Future of Jobs Report 2023 found that nearly three-quarters of companies surveyed plan to adopt AI, with 50% expecting it to lead to job growth. In the book Bridgital Nation, our Chairman, Natarajan Chandrasekaran, predicts that AI could add 30 million jobs by 2025 and provide a major impetus to economic growth.

Unsurprisingly, the impact of AI on the economy will be a major topic at the forthcoming Annual Meeting of the Forum in Davos, Switzerland. And it is easy to see why; some projections suggest the addition of an economic boom equivalent to India and China’s output combined by 2030. We’re at the start of a transformative decade in which generative AI is already reshaping work, life and business. In addition to its business impact, it will unlock benefits in healthcare, education and climate change.

Yet, despite the intense focus on AI, we are missing something fundamental. Public attention has largely been on increasing productivity with accelerating outputs. But the real potential lies in the decision-making processes that precede that. Here, AI can augment what humans are capable of – taking them from average to great, making novices as proficient as experts – rather than rendering them obsolete.

Before looking into this, let’s define what AI means. Its evolution has been rapid and can be delineated by its expanding capabilities. Initially, AI focused on recognition tasks, like identifying objects in images. Its next iteration involved reasoning, analyzing what was happening, why it was happening, the likely outcomes, what we should do about it, and decision-making based on that understanding.

The most transformative shift happened with the advent of generative or operative capabilities, exemplified by Large Language Models (LLMs) like GPT, LaMBDA and LLaMA. These models leverage predictions made during the reasoning stage and can make decisions and propose actions.

LLMs go beyond the information fed to them and extrapolate, resulting in non-deterministic responses to the same input. They have the power to extract insights from unstructured content and when combined with enterprise-specific models, they can create a knowledge superstructure, enhancing decision speed, effectiveness and customer experiences.

It’s clear that the most effective use of generative AI will be in augmenting human creativity, rather than in replacing humans. AI won’t replace jobs, but make people better at them, in part by the democratization of knowledge. This will allow for jobs to be elevated by AI, levelling existing knowledge gaps and increasing global access to information. Indeed, the World Economic Forum predicts that AI will be a net job creator between now and 2027.

While there is an onus on businesses to support their staff in re- and upskilling in line with these developments, the benefits are clear for all to see; increased efficiency and, thus, often higher output are welcome side effects. The need to reskill people must be our first priority towards realizing this opportunity. At TCS, we have trained over 100,000 people on generative AI in the past year, and I am heartened to see similar efforts across the industry, which are essential in building up an AI-ready workforce.

In today's knowledge economy, enterprises generate value through knowledge work, which involves decisions and actions taken by individuals or groups. Traditional techniques like machine learning and data analysis have been used to extract information from structured data. But then there is an organization's tacit knowledge: the unstructured content and data – often acquired by experience, observation, imitation and practice – that cannot be easily codified.

This reliance on tacit knowledge makes decisions challenging to explain and causes variability in decision outputs, outcomes and customer experiences. GenAI or large language models can potentially extract insights from such unstructured content. Foundational models, such as GPT, LLaMA and open-source alternatives, are “world-wise” and can integrate common knowledge that may exist offline, such as in books or paintings.

Combining such models with enterprise-wise ones and traditional AI/ML techniques, a knowledge superstructure can be created within an enterprise, increasing the speed and effectiveness of decisions and actions. This can improve customer experiences, productivity and talent utilization. The greater benefit of these technologies is in digitizing an enterprise's knowledge and reducing reliance on tacit knowledge in decisions and actions.

This requires creating large numbers of enterprise-fine-tuned purposive models or agents for each activity, which will be further augmented by these knowledge superstructures. AI should be viewed as a business redefinition exercise involving various parts of the business, such as legal, security, data, compliance and tech teams, to identify the highest value activities that can be transformed through a knowledge superstructure.

Enterprises will need to invest in a four-tier architecture powered by hundreds or thousands of purpose-built models optimized for cost, quality, security and privacy. This complex undertaking offers significant opportunities to reduce reliance on tacit knowledge, deliver elite quality value, and reduce variance in output quality. A sophisticated and well thought-out strategy is required to drive this transformation within an enterprise.

Ultimately, generative AI offers an enterprise-wide opportunity to improve decision-making processes and overall efficiency. This integration digitizes enterprise knowledge, reducing reliance on tacit knowledge and boosting productivity and talent utilization.

And it’s here where the current focus of public debate is misplaced, focusing almost exclusively on “action”, rather than on the potential to digitize an enterprise's knowledge and reduce reliance on tacit knowledge in decisions and actions. Simply put, the real value of AI and machine learning lies in the step before output, in improving decision-making.

Why? Because it can help us overcome the gap between average and great outputs and unlock higher-quality outputs more consistently. It allows us to democratize knowledge. This could help improve outcomes in many areas – think of enabling doctors to diagnose and treat patients more consistently despite differences in experience and time constraints.

If we harness generative AI correctly and focus on how it can augment our decision-making and creativity, we can unlock a powerful knowledge superstructure, enhance decision speed and effectiveness, and supercharge customer experiences.

Turning average or even good work into great. That is the true opportunity ahead.