Fundamental challenges that SMEs must address before unlocking AI’s potential

Several facets of Artificial Intelligence (AI) including image pattern recognition, Natural Language Processing (NLP), reasoning and pattern recognition have made rapid strides in recent years. The growing popularity of ChatGPT and GPT4 represents a potential tipping point or ‘paradigm shift’ as Thomas Kuhn puts it, as far as the usage of AI is concerned.

Indeed, the progress of technology has largely followed the so-called Gates’ Law, which states that “Most people overestimate what technology can achieve in a year but underestimate what it can achieve in ten years.” Gates himself was likely paraphrasing the words of Roy Amara, a Stanford computer scientist who first outlined this philosophy way back in the 1960s.

However, in our experience of working with several customers in this space, including a large multinational FMCG player, the extant requirements of clients are often more mundane. As one goes down the pecking order in terms of company size, for example, if one were to consider Small and Medium Enterprises (SME), these ‘mundane’ problems become even more acute.

We list some of these ‘simple’ and straightforward challenges as follows:

• Many enterprises wish to deploy Machine Learning (ML) solutions but lack knowledge of how effectively to deploy them, which includes the pre-requisites such as volume, relevance, and real-time data. As technology advances, so do the expectations of clients and customers. The potential for disappointments is directly linked to the growth of expectations. Given that trends and technology change quickly, the requirements of companies may need modifications, but firms may not be aware of them; and the earlier requirements aren’t quite feasible or are outdated in the current market scenario.

• The rapid changes in technology have not been an unmixed blessing as some of these technological changes have led to rising IT costs, without there being adequate budgets to effectively support these.

• Many enterprises face challenges in collecting data and harnessing its full potential. An incremental approach often fails to pay off in terms of costs v/s benefits and a full-scale implementation, though bringing more benefits is also more expensive. In addition, some of the costs are front-loaded but the benefits accrue over a longer time frame.

• Another interesting corollary to this is that some enterprises use too many analytical tools or techniques but are unable to utilise the insights so generated adequately as they do not have manpower of sufficient skills or numbers to utilise or even see those insights when deployed. Existing IT manpower must keep itself abreast of the rapid growth of AI-based tools and their effective usage, sometimes a tough task for many companies.

• Enterprises are looking for ready-made “plug and play” type solutions; IT firms need to find a way to create specific solutions in collaboration with clients-this may be achieved using POCs. POCs can turn out to be a win-win as expectations get aligned over time about the product and a customized solution specific to customer needs is created.

In our experience, the rising capabilities of technology will deliver on its potential only when such aspects are addressed together.

Rather than investing in the development of Machine Learning algorithms per se, some providers have started to focus on working to develop ML use cases that will provide the most cost-effective returns to customers. Such solutions need to be made case-specific and industry-specific, rather than adopt a ‘cookie-cutter’ approach.

In addition, practices such as the adoption of microservices / composable architecture allow customers to replace/upgrade only a certain part of their legacy solutions or technology that is needed. This focused and low-investment approach helps companies achieve a competitive edge in the required areas without changing the entire IT infrastructure.

A failure to address some of the ‘mundane’ matters listed above explains why many digital transformation or AI-led initiatives fail or are seen to have delivered less than what was promised. Taking care of such issues by seeking the right kind of help could therefore make a big difference in achieving the promised results.

About the authors:

Sidharth Balakrishna, an IIM Calcutta and SRCC alumnus, has led digital transformation projects in many companies and is also an Adjunct Faculty at IIM Indore and SRCC.

Amit Verma is the CEO of DigitUp, a mid-sized IT solution provider.

Disclaimer: The views expressed in this article are those of the author and do not necessarily reflect the views of the Economic Times – ET Edge Insights, its management, or its members

Scroll to Top