Is Generative AI overhyped?

Artificial intelligence (AI) and generative artificial intelligence (GenAI) in particular, have captured the imagination of the world with an ability to create new content, including text, images, videos and code, as well as interpretations of existing data and content. Amid all the excitement around the potential and massive investments being made in AI, more and more questions are starting to surface as to whether the hype surrounding GenAI is justified, or if we are naively and vastly overestimating its potential?

Global spending on artificial intelligence (AI), related applications and other services is estimated to almost treble by 2028, according to the International Data Corporation (IDC) Worldwide AI and Generative AI Spending Guide – and a significant portion of that is attributable to GenAI. The IDC forecasts an increase in AI spend from $235 billion to $631 billion between 2024 and 2028 as a result of the massive and ongoing global advances in AI and GenAI.

A global McKinsey survey conducted in 2024 found 65% of respondents reported that their organisations are regularly using GenAI, nearly double the figure from just 10 months ago, and with 75% predicting that GenAI will lead to significant or disruptive change in their industries in the years ahead.

GenAI enabling cost reductions and revenue increases

There is no shortage of surveys published in the past 12 months that clearly report on how organisations are already seeing material benefits from GenAI use, including cost reductions, productivity and efficiency gains, and revenue increases. The survey findings suggest that organisations are using AI in a far wider array of business functions than in previous years, with GenAI adoption most common in functions such as marketing, pre-sales, sales, product and service development and across internal IT.

In very simplistic terms, GenAI is artificial intelligence capable of generating text, images, videos or other outputs, usually in response to natural language prompts, and from a chatbot perspective, mimicking human conversation.

Chances are that if you are reading this article, you have come across some of the most common GenAI tools in use today, including OpenAI’s ChatGPT, Google’s Gemini, Microsoft’s family of Copilots, Midjourney and Anthropic’s Claude, to name a few.

GenAI the number one type of AI solution being deployed

According to Gartner, GenAI is the number one type of AI solution currently deployed in organisations, with 29% of respondents saying they have already deployed and are using GenAI. Interestingly, only 9% of companies regard themselves as AI-mature, where AI-mature organisations are defined as typically having a scalable AI operating model, balancing centralised and distributed capabilities; focus on AI engineering and have a systematic way of building and deploying AI projects into production; having invested in upskilling and change management; and with a focus on trust, risk and security management capabilities to mitigate the risks that come from AI implementation.

AI maturity is not a prerequisite to business value generation, however. Gartner’s research indicates that early adopters of GenAI are reporting a range of business improvements including 15.8% revenue increase, 15.2% cost savings and 22.6% productivity improvements on average. There is still much work to be done on matching business value, direct return on investment and future value impact of GenAI utilisation, but this initial feedback is quite clear. In answer to a frequently asked question around the monthly costs associated with a number of these tools, we have developed a ROI calculation tool – and we have shown that even for many junior employees, there is a breakeven on these costs, whereas for the most senior levels of the organisation (CxOs), ROI can be in excess of 700% for this monthly subscription spend.

GenAI’s benefits needs to be countered by its limitations

While GenAI certainly offers significant benefits, it also comes with limitations. The most significant limitation is that GenAI models still produce inaccurate responses. Over time, these inaccuracies and biases can be trained out of these tools, and we have already seen significant improvements over the past two years in this regard. These optimisations and improvements will only increase over time.

One of the least-understood aspects of the GenAI large language models (LLMs) that are the foundation for natural language query interfaces (eg, ChatGPT and Gemini) is that there is in fact limited intelligence behind these interfaces – and the generated outputs are actually the results of a probability calculation, rather than a human-style understanding of the question or prompt.

It is for this reason that I love the definition of AI to be “simulated human intelligence”, as it is currently a simulated behaviour of what could be expected from a human in terms of the generated output, rather than an actual intelligence per se.

The societal implications of GenAI can also not be ignored. As GenAI becomes more sophisticated, there is a very strong likelihood that some employees doing mundane and repeatable work will become redundant as their tasks become automated – and this trend has already started. While AI will result in some new jobs being created, unfortunately these will not make up for the inevitable job losses that will continue to increase more and more over the next decade. Apart from many of these jobs that can be fully automated, the balance of jobs for the next five to 10 years will most likely be humans working with the assistance of AI. There is real substance to the quote that “for the next decade, most roles are not at risk from AI, but from a human who knows how to use AI better than you do”.

Adopting GenAI needs to be adopted ethically and responsibly

The need for governance around AI initiatives, solutions and adoption cannot be ignored. As the World Economic Forum has pointed out, along with AI’s potential value, there are concerns about its risks, including bias, safety, security and loss of reputation if something goes wrong. Adopting the technology ethically and responsibly is therefore fast becoming a necessary consideration for business leaders.

The United Nations believes it should be taking the lead on creating a global framework for AI governance, and its newly released report, Governing AI for Humanity, is the first global approach to addressing the governance challenges posed by AI. At Altron Digital Business, we have already applied our minds to this challenge and have developed an AI governance framework that addresses way more than just ethics, TRiSM, explainability and the currently trending areas of AI governance focus. We will keep this updated as global initiatives progress in this regard, but for now, we believe we are leaders in the AI governance journey.

It is important to note that locally, the South African government has recently released a draft AI Policy Framework for comment. This lays the groundwork for a national policy, AI regulations and potentially an AI Act to guide the responsible and ethical development and utilisation of AI across all industries in South Africa.

There is no going back: GenAI has forever changed how we work

Despite the hype around it, GenAI changed the face and productivity of knowledge work – forever. Consider, for example, how the nature of work has changed in just three decades. From the early 1990s with no public internet or mobile phones (and limited adoption to e-mail), it’s hard to imagine the world of work without these critical tools. In the same way that the internet, e-mails and smartphones changed the world of work, GenAI is delivering a similar step change, offering greater efficiencies and unprecedented productivity gains.

There is no going back. Those who ignore GenAI will fall behind – in some cases, critically. And to ensure that our next generation of employees can utilise these tools for maximum benefit, education curriculums need to start including AI-assisted work modules and teaching learners how to use the technology responsibly, while bearing its limitations in mind. The ability to prompt and curate output are two skills everybody will need to have in the workplace, and the sooner learners understand that how they ask a question can result in optimal or sub-standard output, the better. Rather than instilling fear about GenAI in the next generation of employees, we should be ensuring that they are skilled at using AI to perform their jobs.

The world of work has turned a corner – enjoy the ride as GenAI hype morphs into your future reality.

This article was first published on ITWeb