IBM UK CEO on AI Adoption, Skills and Workplace Productivity

For UK businesses, the AI adoption gap is already visible in deployment figures. IBM’s Global AI Adoption Index found that around 30% of UK enterprises with more than 1,000 employees had actively deployed AI, compared with almost 60% in India, 58% in the UAE, 53% in Singapore and 50% in China. IBM UK and Ireland chief executive Dr. Nicola Hodson sees room for that gap to narrow: about 41% of large UK enterprises were exploring or experimenting with AI, a pool of potential deployments still to come.

Hodson’s argument is that the next step depends less on enthusiasm for AI than on whether organizations can put the right skills, data and controls in place. She expects AI tools to change tasks and create room for higher-value work, while stressing that businesses need a clear reason to adopt them and a way to evaluate whether they deliver commercial value.

From experimentation to deployment

The Index figures suggest that UK adoption is at an earlier stage than in several other surveyed markets. The 41% of large enterprises experimenting with or exploring AI points to continued activity, but moving a test into production requires decisions about cost, data access, workforce readiness and how outputs will be checked.

Hodson identified three main barriers for enterprises: shortages of skilled people, the cost of implementation and the complexity of business data. She added governance as another requirement: AI systems should produce accurate results, avoid bias and toxic output, and make it possible to understand what data informed a result.

These challenges are connected. A company may have a promising use case but lack staff who can assess the system, data that is organized and accessible, or a practical process for monitoring its output. That makes adoption a business and operating question as much as a technology choice.

For skills, Hodson argues that organizations should train staff to use AI now. Teams that know how to apply the tools may be better positioned to get value from them. Where internal expertise is limited, an experienced partner can help identify suitable use cases and develop a plan for workforce training.

IBM says it has trained more than 1,000 consultants to form its Center of Excellence for Generative AI. The group works with clients on strategies to build employee skills and assess where generative AI may have commercial value. IBM’s client engineering team also works with organizations on pilots, helping them test possible applications and decide which are worth extending.

That approach puts evaluation ahead of broad rollout. A pilot gives teams an opportunity to learn how a system behaves in their own work, whether staff can use it effectively and whether the result justifies the cost. Hodson’s advice is to identify a business goal first, then test a small number of use cases before moving into wider deployment.

Data and responsible use

Data complexity is a separate obstacle, though it can determine whether an AI project works at all. As Hodson put it, “AI is as good as the data it’s working across.” She recommends building an organized, secure data platform and describes hybrid cloud as a way to connect data and services spread across different cloud environments. A company needs a workable view of its data wherever it sits to run and manage AI across the organization.

Trust is another adoption concern. Almost 25% of UK respondents to the IBM Global AI Adoption Index said ethical concerns were a barrier to using AI. Hodson pointed to familiar problems: outputs that are wrong, inappropriate or potentially harmful. For business use, she said, systems need to be accurate and trustworthy, with explanations of how they reached decisions.

IBM launched watsonx, an enterprise AI and data platform, with a governance layer intended to show where model data comes from and which data informs outputs. The platform can flag inaccuracies or bias and automate checks against relevant regulations and company rules. Hodson compared this transparency to an AI “nutrition label,” making it easier for organizations to inspect the basis of a model’s output.

IBM also says it established an AI Ethics Board to review products against the company’s ethics and trust principles before release. Such processes address one part of responsible deployment; organizations still need to consider how a tool will be used in their own workflows and how people will review its results.

Training for work that changes

Hodson expects AI to reshape tasks rather than simply remove work. She recalled a discussion about the washing machine: a technology expected to free up time did so, but people found other tasks to fill it. Her analogy for software development is similar. If AI makes a developer 30% more productive, she said, that does not necessarily mean the developer does 30% less work. The time could instead go toward more valuable tasks.

In her view, higher productivity can also help businesses grow and create roles. “Many jobs will evolve as we increasingly see AI and humans working hand-in-hand,” she said, adding that new roles may emerge that companies have not yet anticipated. That outlook still puts pressure on workers and employers to build skills as job requirements change.

Hodson cited an estimate that around two million jobs went unfilled in the UK last year, underscoring the scale of the skills challenge. IBM’s SkillsBuild program aims to train 30 million people globally in technology skills by 2030. Its offer includes an AI Fundamentals course available through an online portal. IBM has also committed to training two million people globally in AI skills by 2026, working with schools, colleges and universities.

The company says it works with the National Cyber Security Center on a cyber-first program that includes education for girls. Hodson cited work focused on girls in Manchester and said the program involves more than 1,000 schools. These initiatives reflect a wider need to make technical training accessible to people who may not have a conventional route into the field.

Generative AI may lower some barriers to using technology because people can interact with systems through ordinary language. “Coding is now in natural language, in plain English,” Hodson said. That can make it easier for non-specialists to try tools without first learning a programming language, though being able to prompt a system does not remove the need to check what it produces.

She gave two workplace examples. A consultant can use a tool to draft user personas—profiles of the people a service is intended to serve—in minutes, a task that previously took days of research. IBM also uses AI in Ask HR, its employee service. Staff can submit plain-English requests and receive responses quickly; Hodson said first-contact resolution and satisfaction are high, without giving specific figures.

Where companies are putting AI to work

Hodson grouped common enterprise uses into three areas: customer service, digital labor that automates repetitive manual tasks, and application or code modernization. The examples she described range from handling routine requests to helping employees work through large volumes of information.

At University Hospitals Coventry and Warwickshire NHS Trust, IBM Consulting worked on a system intended to reduce missed appointments. The reported no-show rate fell from 10% to 4%, enabling doctors to see more than 700 additional patients each week without extra staff. IBM estimates the solution could reduce the trust’s waiting-list backlog by 10% to 15%.

For NatWest customers, the bank’s digital assistant Cora is being enhanced with generative AI from watsonx. The stated aim is a more conversational assistant that can provide greater value to customers. IBM also worked with The All England Lawn Tennis Club on AI-generated commentary for match-highlight videos on Wimbledon’s digital platforms, an example of using generated content to extend digital audience experiences.

In Norway, IBM worked with a media company on a generative AI tool to help investigative journalists research and prioritize stories. The intended benefit was to reduce time spent on preparatory work, leaving more time to report and produce stories. These cases illustrate distinct kinds of value, but the article does not provide comparable measures for the assistant, commentary or media applications.

Hodson said 60% of early adopters that had overcome adoption barriers and deployed AI were making further investments. Their willingness to spend again suggests they had seen benefits, though the figure does not specify which uses drove those decisions or how gains were measured.

Productivity claims and practical limits

IBM’s own examples offer a sense of the tasks it expects AI to absorb. Hodson said generative AI can summarize lengthy reports and meetings, draft correspondence, support research and create presentations. She expects assistants to take on more of this work, potentially returning weeks of time to professionals over a year. The amount will depend on the task and how much human review remains necessary.

Within IBM, the company’s HR team saved 12,000 hours over 18 months by automating processes that previously required repeated exchanges between managers and employees. IBM Consulting also reduced its promotion cycle from 10 weeks to five by automating administrative work. These are company-reported results, and the source does not break down the systems’ costs or the exact methods used to calculate the savings.

For IT modernization, IBM’s Code Assistant can convert older COBOL code into Java. Hodson said this can reduce costs and speed modernization, particularly because experienced COBOL specialists are expensive. The source provides no conversion accuracy or deployment figures, so the practical value will depend on how much human review and testing a migration requires.

AI-enabled software may also help reduce cloud operating costs and energy use through ongoing monitoring and data analysis, Hodson said. In marketing and communications, assistants can draft initial content and generate ideas, giving teams more time for strategy and creative decisions. In both areas, the proposed benefit comes from automating routine analysis or drafting; the quality of the output and the work required to review it remain relevant to the business case.

A measured path to wider use

Hodson’s advice to UK businesses is to make data accessible, train employees and choose use cases tied to a clear commercial goal. Hybrid cloud can help bring together data held across different environments, while pilots give teams a way to test whether a particular application is useful before expanding it.

She expects more AI assistants to work alongside employees, which makes preparation a continuing task. Businesses will need to help staff learn where the tools fit, how to judge their output and when a person should take over. The adoption figures point to interest across the UK’s large enterprises; the next test is whether organizations can turn that interest into deployments that are useful, well governed and worth maintaining.

For Hodson, AI’s workplace role is collaborative: automate selected tasks, give people more time for work that needs judgment, and create new kinds of work as organizations change. The near-term measure of progress will be practical—whether companies can solve data and skills bottlenecks, demonstrate value in focused applications and earn enough trust to use them more widely.