Implementing GenAI remains challenge for majority of firms

The excitement around generative AI continues to struggle to manifest genuine gains for businesses, according to a new report from Capgemini. While almost two-thirds of organisations have piloted initiatives with the technology, three-quarters say that large-scale deployment across their operations would be very difficult.

The importance of generative AI has risen steadily on the corporate agenda, over the last 18 months. The hype around GenAI, and its purported potential to reduce costs and boost efficiency and output has led to its fast-tracked adoption by businesses across the industrial spectrum – with one poll in mid-2024 suggesting 62% of respondents thought it would be crucial in the coming three years. Meanwhile, 57% said they were at least exploring use cases, and 58% were already piloting initiatives involving the technology.

But amid the sustained hype of what the technology could do in the future, its practical uses in the present remain muted. Another recent survey found that even among industry pacesetters, only a third of firms have actually completed an AI project – while the completion rate for AI projects across all corporate functions and departments stood at a meagre average of 26%.

Delving into why this might be, a new report from Capgemini Research Institute has highlighted a surge in organisations attempting to unlock value with the technology – but struggling to realise notable financial gains from it. According to the report, 60% percent of organisations have implemented pilots or early proofs of concept (PoCs) for GenAI initiatives using their enterprise data – and yet an even higher 75% of organisations say that large-scale deployment of GenAI PoCs is a significant challenge.

The rate struggling to deploy GenAI more widely rises and falls according to industry. The worst-performers were found in the utilities and automotive sectors, where 84% of executives said scaling GenAI was a “major challenge”. However, no sector saw that rate fall below six-in-ten, with the best performers coming from the retail and life sciences sectors – where 64% said GenAI scaling was a “major challenge”.

This may be indicative of those sectors being better-versed in leveraging data to adapt their business offering, or create new products. Indeed, the researchers suggested data may be the key to unlocking GenAI’s potential – noting that while 59% of businesses believe their technical foundations for data, technology and infrastructure are strong, only 40% of data experts agree.

At the same time, 87% of data executives said their organisations did not use external data to train GenAI, while 55% added they were tapping into structured, semi-structured, and unstructured data for data-powered decision-making and implementation. This means that many firms may be limiting how much data their GenAI can access to learn with – but also may be feeding it information which could be irrelevant to its designated tasks.

Commenting on what may need to change, the researchers noted, “Ensuring the reliability and correctness of input and output data is critical to the success of generative AI. However, only 49% of data executives agree that their organisations are well prepared in this respect. There is a perception gap regarding transparency of data, with 87% of business executives believing their organisations are prepared on this front, whereas only 37% of data executives agree. As many as 61% of business executives cite the accuracy and impartiality of input and output data as major concerns.”