Eighteen months since the launch of ChatGPT, the expectations of Gen AI transforming businesses and societies remain just that – expectations! 

Companies caught on the hype have committed to spend around ~$1tn in data centres, chips, and other AI infrastructure to ride the gen AI wave. While the investments continue unabated, the returns are nowhere in sight. Now industry analysts have started to publicly question if the huge investments in GenAI will ever pay off!

The biggest cause for concern is the fact that there is no major transformative application for gen AI yet. The most clear-cut use case of gen AI has been in making existing processes, such as coding, more efficient. But even here, the cost of using gen AI is much higher than that of the incumbent non-AI methods. Most accepted gen AI products, such as chatbots, lack clear monetisation strategies as well.

According to Gartner’s hype cycle, gen AI has gone beyond the peak of inflated expectations into the trough of disillusionment. By 2025, 30% of gen AI projects will not progress beyond proof-of-concept. And only one in four AI-powered automation projects will actually be cost-effective. In the US, the overall productivity gains from such automation initiatives will be less than 1% over the next ten years.

Why do gen AI Projects Fail?

The reasons for gen AI projects failing or not delivering value are manifold.

Lack of access to adequate data, poor quality of available data, and unclear business value derail many AI projects. Even if these resources fall in place, the high costs make most projects unviable. Chip shortages and huge energy demands make projects fail the cost-benefit analysis.

Also, the public is becoming more distrustful of AI in general and gen AI in particular. Concerns related to data misuse, privacy, and job losses have always accompanied AI adoption. A growing concern now is algorithmic biases and lack of accountability.

Many promoters position AI as a neutral tool. But algorithmic bias makes gen AI tools far from neutral. Most AI algorithms being “black boxes,” with no clarity on the rationale behind the answers, increases distrust. To make matters worse, gen AI systems and their promoters do not take accountability for errors and mistakes.

On the corporate front, the extreme overvaluation of companies involved in the AI space indicates a bubble. A bubble develops when future cash flows do not support expensive valuations. Many analysts see big leaps in stock price without the companies generating corresponding real value to back it up as a bubble waiting to burst. The market and wider economic conditions today resemble the conditions that led up to the dot-com bubble burst of 2000. Both periods saw huge increases in stock prices and valuations. Hopes and expectations rather than tangible results drove the increase.

Increasing regulation related to AI safety, ethics, and data privacy could also slow down the momentum of gen AI development. Anti-trust lawsuits and verdicts on genAI models violating copyright could spell further doom.

Gen AI failure can be costly for enterprises. Gen AI projects can cost millions of dollars to implement and incur huge ongoing costs. For instance, rolling out a gen AI virtual assistant costs upward of $5 million and incurs a recurring expenditure of above $8,000 per user per year.

The Hope for Gen AI

But it is not all doom and gloom for gen AI. Despite not meeting the inflated expectations, Gen AI does deliver significant benefits. Operating profit gains from AI doubled to almost 5% between 2022 and 2023, with the figure expected to reach 10% by 2025.

The talk of AI not making money is the opposite end of the hype.  GenAI is in the early stages of market evolution. At this stage, the wider market ecosystem favours supply-side revenue generation. The demand-side enterprise adoption remains nascent.

The top companies investing in AI, such as Google, Meta, Microsoft, and Nvidia, have strong fundamentals. These companies are profitable and have multiple revenue streams. They already have huge user bases and vast amounts of proprietary data related to AI. Even without an AI-doinated society coming to pass, these companies can profit from genAI. They already have a clear monetisation path. For instance, advertising services, personalisation, online shopping, and more already deliver considerable profits.

The cost of new technologies tends to fall over time, and the gen AI will follow a similar pattern. When the infrastructure and platforms fall into place, killer apps will follow.

Focus on Delivering Business Value

CIOs can fuel the next wave of IT operations by ignoring the hype and focusing on the potential of gen AI.

Easy wins and the obvious no longer suffice. CIOs need to look beyond digital assistants and intelligent chatbots. They need to:

They would do well to:

• Fine-tune existing AI models for specific use cases before investing in training models from scratch.
• Focus on custom applications that deliver tangible business value.
• Continue to encourage experimentation and innovation. The real value creation of gen AI is in the long term. A culture of continuous learning and innovation builds internal AI capabilities. Identifying new enterprise-relevant use cases becomes easier in such a culture.

Determine ROI for Gen AI Investments

Determining the ROI for gen AI uses is difficult. Many benefits, such as improved customer satisfaction and better decisions, are intangible. The benefits may not realise immediately either and may manifest only in the long term.

A key challenge is quantifying the efficiency gains delivered by gen AI in monetary terms. Consider a virtual assistant. Using a gen AI-powered virtual assistant to compose emails may free up the employee to perform higher-value tasks. But most enterprises have not tracked the time employees spent on composing emails in the first place, to quantify the gains.

CIOs need to:

• Quantify the impact of gen AI wherever possible. Compare the revenue increase, operational cost reduction, savings in labour costs, and more.
• Conduct customer satisfaction surveys to gauge the impact on the customer.
• Focus on qualitative metrics, such as tracking the new products or services enabled by Gen AI.
• Benchmark the gen AI impact on industry standards.

Shun the Gen AI or Bust Approach

Many enterprises caught in the gen AI hype adopt a solution in search of a problem approach. The right approach is to first identify problems that exist in the enterprise. Next, consider all viable technologies to overcome them. Consider gen AI as one of the several possible solutions for the issue. Several non-AI and even other AI solutions are cheaper than gen AI and deliver much better value.

CIOs adopting a gen AI or bust strategy are more likely to go bust.

As practical realities temper unrealistic expectations, the gen AI market is in the midst of a churn. But it is by no means a burst. Used the right way, gen AI delivers significant value addition. Developing business cases, upgrading processes and taking care of technical debt takes time. Once these things fall in place, gen AI will start delivering a transformative impact for the enterprise.