Disruption and doubt: How UK IT leaders view GenAI
Computing research finds most still taking a cautious approach
Fifteen percent of 100 UK IT leaders polled by Computing in September said they were using GenAI in an organisational capacity (as opposed to ad hoc experimentation).
This compares with 10% who said the same in February.
A further 17% were rolling out proofs of concept, the same proportion as in February, while a hardcore 31% said they had no plans, again a consistent result between the two surveys.
At the same time, even as more use cases are being identified, there is apparently increasing scepticism about whether GenAI will live up to its promises.
Asked whether there could be a drop in business interest in AI in the next 18 months, 38% said yes, compared to just 17% in February. A quarter (25%) believe most AI announcements to be “snake-oil”, up from 15% in February.
This, no doubt, is partly a response to the increasing pressure from management to adopt AI, and a feeling that its shortcomings are being overlooked (57%); that privacy, bias and security concerns need addressing (54%); that promised benefits will fail to emerge (43%); and that AI will face road blocks in terms of its scalability, not least because of its resource requirements (40%).
“Like all these new 'wonder cures' implementing and setting it into our culture will be complex, costly and long-winded, but we will, eventually, get where we need to be,” said an IT manager in higher education.
“Looking at net zero and the need for global sustainability, the power consumption required is simply outrageous,” added a CTO in the service sector; while the director of a cybersecurity institute said that it can be wise to wait until the irrational exuberance phase has run its course: “Genuine worthwhile benefits and applications will emerge after the downturn.”
In comparison with earlier forms of AI, the new kid on the block lacks reproducible business use cases, meaning that some sort of market correction is inevitable, said a consultant: “Predictive AI is solid, but GenAI is overhyped and it underperforms.”
Which sectors will first be disrupted by AI?
Hyped or otherwise, GenAI is already making its presence felt - but in some sectors more than others. The tech sector is first in line for disruption, according to our respondents (54%), followed by media/marketing/PR/advertising (52%) and education (34%).
Interestingly, the public sector, where the government has put significant store in using AI to improve productivity, was towards the bottom of the disruption rankings, although health and education (generally part of the public sector) both placed higher.
“Software development itself will be most impacted as it lowers the bar for new product development,” stated the director of an IT consultancy.
“The advances are in generative AI, so areas where lots of copy and content can be generated automatically be impacted,” said the software director of a business consultancy, who predicted that the business services sector would among the first to feel the impact.
“As ever the areas that need it least will be the first to adopt and the first to fail” offered an IT manager in education. “The slower, more measured approaches by the possibly 'wiser' areas such as education and local government will do far better in the medium to long term.”
When managing the impact of new technologies, establishing controls is key said an IT service manager in technology. “Disruption may be positive rather than the assumed negative. With governance, positive outcomes are possible.”
Risks and rewards
The main positive outcomes hoped for by respondents were increased productivity, finding new ways of doing things, and reducing drudgery.
“In my industry in particular we should see an increase in the volume of scientific research and improve the speed at which this can be validated and published,” said the head of DevOps in a scientific organisation.
Here too though, there responses were characterised by a certain wariness. “It could replace repetitive tasks but not original thinking,” said an operations manager in the services sector.
Use cases for GenAI
While AI promises to unleash some revolutionary new use cases, currently organisations are mostly investigating its potential for automating existing ones. Just 14% said they were looking at new use cases over the next three years, versus 75% who plan to apply AI to existing tasks.
For 5% of respondents, AI had already created new jobs. These were mostly in training and research.
Implementing GenAI in the enterprise faces several challenges, not least a lack of skills (mentioned by 44%). The difficulty of integrating AI into existing systems (46%) and ethical and legal concerns (40%) were also thought to be barriers to adoption.
The rapid development of the field and the absence of readily applied governance frameworks has made some wary of adoption, with perceived organisational and societal risks including worsening bias, discrimination, disinformation and propaganda (mentioned by 60% and a concern with GenAI); misuse by malicious actors (59%); and security and privacy concerns (44%).
“As long as we, and other slower adopters, learn the many lessons that will come from the early adopters we should be OK, but 'slowly-slowly-catchee-monkey' will always be our watchword, and no big projects are planned yet,” said the IT manager in higher education.
Too big to fail?
One of the risks mentioned was market capture by large corporations, which are now the only ones with sufficient resources to create competitive LLMs from scratch.
Half of September’s respondents (50%) believed that “Gen AI will centralise even more power in the hands of tech giants” compared with 45% in February.
The CrowdStrike incident in the summer resurfaced concerns about being dependent on a small number of providers, and recent antitrust moves by the US authorities have focused minds on big tech companies being, effectively, monopolies.
This is a complex area, but a majority (52%) of respondents felt that some of the largest companies should be broken up.