AI may be a bubble, but we're forging ahead anyway: research
Computing research finds sentiment growing warier, but little sign that that is slowing activity
Computing’s latest audience research reveals that while concern over the financing of the AI boom may be increasing, few IT leaders are allowing that to slow their plans.
Just under three in ten of 110 respondents (28%) polled in June believe AI is a bubble, while 39% disagree and 33% remain undecided.
Compared to January, when we previously asked this question, the number who buy into the bubble hypothesis has barely shifted, whereas the proportion rejecting it has fallen from 50% to 39%, with more people now saying they’re unsure. Not a massive movement over the six months, perhaps, but there appears to be growing uncertainty about the technology’s trajectory.
However, concerns over bubble economics, such as they are, are not translating into retreat. As with previous technology booms, IT leaders appear willing to accept the possibility of even quite severe market upheaval in the belief the underlying tech will eventually deliver.
Bubble sentiment may be growing, but not enough to deflate expectations
The reasons cited by those who believe AI to be a bubble include a lack of clear business models, concerns about circular financing between major vendors and the absence of compelling productivity gains. Respondents also point to infrastructure constraints and the potentially disruptive impact of cheap open-source alternatives.
“Too many players, not all ideas will prove viable/profitable. Inflated valuations affecting venture capital costs, unsustainable levels of investment,” posted a strategist in government.
“I feel currently AI is being touted as the solution to everything,” said an IT systems manager in education, adding: “This technology clearly has a place going forward but it is sometimes used where it compromises effectiveness or the requirement for oversight and validation of output is neglected.”
Overall, however, the findings suggest that IT leaders distinguish between the technology and the market forces driving its momentum. It’s possible to be sceptical about valuations, vendor hype and investor expectations while remaining optimistic about the long-term usefulness of AI itself.
In short, respondents may believe there is an AI bubble, but not that AI is vapourware.
AI deployments: expanding but immature
As we have covered in previous articles on this research, AI adoption is progressing, but most organisations are still some distance from maturity.
Data readiness remains a major hurdle. More than half of the respondents described their data as not ready or only partially ready for AI initiatives, while fewer than one in five considered it to be genuinely AI-ready. Without clean, high-quality, well-governed data, even the most sophisticated models and platforms will struggle to deliver consistent value.
Governance is another evolving field. Most respondents have basic AI policies in place, but relatively few (22%) have established formal governance structures and only a small minority (6%) reported comprehensive board-level oversight.
In terms of models and customisations, most organisations are keeping it simple, relying on off-the-shelf models or prompt engineering rather than more advanced approaches such as retrieval-augmented generation (RAG) or fine-tuning.
Evaluation practices are similarly immature, with many organisations relying on informal testing, and relatively few measuring progress through clear business KPIs and benchmarks.
Together, these findings suggest that AI adoption is broadening faster than organisational maturity. It’s wide and shallow.
AI goals: let's get the fundamentals right first
Respondents' priorities for the next 12 months support the picture of organisations still focused on the foundations rather than transformation.
The most common goal, by some distance, was experimenting with new use cases (66%). This was selected by two-thirds of respondents, significantly ahead of scaling existing deployments (32%).
Other leading priorities include improving governance (38%), building internal capabilities (38%) and reducing costs (32%). Notably, measuring ROI and reducing costs were important but secondary concerns.
Many organisations are still in the discovery phase of AI adoption. Rather than moving toward enterprise-wide deployment, they are still trying to identify where AI can deliver genuine value, and ensure they have the skills and controls needed to scale up.
AI skills: a major barrier
If there is one obstacle that consistently appears throughout the findings, it is skills.
The largest shortages are in AI/ML engineering and architecture, but governance and compliance skills are close behind. This indicates that businesses recognise that successful AI programmes require more than just data scientists and machine learning engineers. They also need people capable of understanding and managing regulation, ethics and risk.
Most organisations are focusing on upskilling existing employees rather than hiring new talent. Partnerships with vendors and consultants are common, but internal capability-building remains the preferred strategy.
This could reflect both the scarcity of specialist AI expertise and a growing recognition that AI skills need to be distributed throughout the organisation rather than concentrated in a small technical team. It also points to a continued lack of opportunity for new entrants.
AI risk
While they may be enthusiastic about its potential, organisations are also mindful of the risks.
Security vulnerabilities, hallucinations, privacy breaches and regulatory compliance concerns dominate respondents' risk rankings. Nearly three-quarters report being moderately to extremely concerned about AI-related business risks. Bearing in mind that this research was conducted before the headline-grabbing “sandbox breakouts” that number will likely have increased.
There is also significant awareness of broader societal issues associated with AI. Respondents expressed concern about misinformation (37%), the blurring of fact and fiction (40%), job displacement (36%) and bias and discrimination (30%).
“Increased reliance and reduced human intelligence,” worried a CTO in retail. “A new generation is growing up who can't listen, learn, understand or think, and rely on AI to do all those things for them.“
At the same time, more than four-fifths say regulations such as GDPR and the EU AI Act are important to their AI strategy.
Forty-four percent expressed themselves concerned or very concerned about the environmental impact of AI, with 16% saying they limit their usage because of this and a further 22% saying they include environmental factors in their evaluations.
The spectrum of opinion that appears markedly more mature and nuanced than some of the public narrative around AI might suggest. Rather than pursuing AI at any cost, organisations are increasingly treating it as a governance challenge requiring oversight, accountability and careful risk management.
Overall, the results reveal organisations neither caught up in AI mania nor throwing up their hands in the face of predictions of imminent collapse. Instead, it shows businesses proceeding with cautious determination. Confidence in the AI boom may be lessening, concerns about a bubble are growing and questions about ROI persist. But investment continues because most respondents appear to believe the technology has genuine long-term potential.
This is the third in a series of articles on the current state of AI usage based on our research among UK organisations. You can read about the cost pressures as usage increases and vendors change their pricing policies and how organisations are really using AI.