AI adoption is accelerating across the UK.
The proportion of UK businesses with ten or more employees using at least one AI technology increased from approximately 12% in late 2023 to 35% in June 2026. However, the Office for National Statistics describes this adoption as “relatively shallow”. Among businesses using AI, the average number of AI technologies used increased only modestly, from approximately 1.4 to 1.6.
Only 10% said they were using AI extensively.
This reveals an important distinction.
More organisations have AI tools. More employees are experimenting with them. But access alone does not mean that AI has become a meaningful part of how the organisation operates.
The next phase of AI adoption will not be defined by how many people have a licence. It will be defined by whether organisations can turn AI into better work, stronger services and measurable value.
Access is an important first step, not the strategy
Giving employees secure access to generative AI can create immediate benefits. People can draft documents, summarise information, explore ideas, analyse text and reduce the time spent on routine tasks.
It also gives employees an opportunity to learn what the technology does well, where it struggles and how it could support their roles.
But access is only the foundation.
When AI adoption remains entirely individual, the value depends on each employee’s confidence, judgement and ability to write effective instructions.
One person may develop a useful approach while colleagues continue working as before. Another may repeatedly create prompts for a task that could have been configured once and shared across the team.
Useful experiments take place, but the learning remains fragmented. The organisation gains activity without necessarily building capability.
This is why usage should not be mistaken for transformation. An organisation can have thousands of AI interactions without materially changing a single client service, knowledge process or operating model.
Productivity and transformation are not the same thing
Most organisations begin with productivity because it is accessible and relatively easy to understand.
The latest UK government research found that increasing efficiency or productivity was the most common reason for adopting or expanding AI. Although most businesses using AI reported some improvement in workforce productivity, more than three-quarters had not yet experienced a change in revenue.
There is nothing wrong with using AI to save time. A faster first draft, a more concise summary or a quicker route through a long document can all be valuable.
The problem arises when these isolated efficiencies are treated as the end goal.
Deloitte’s 2026 enterprise AI research found that 66% of surveyed organisations reported productivity and efficiency gains. However, only 34% were using AI to deeply transform products, services, processes or business models. A further 37% remained at a surface level, with little or no change to existing processes.
Productivity helps people complete existing work faster. Transformation asks whether the work should be organised differently in the first place.
Why AI adoption often remains shallow
Several common patterns keep organisations at the access stage.
The technology arrives before the business problem
Many AI initiatives begin with a tool and a general instruction to find useful ways of using it.
That can encourage experimentation, but it also produces long lists of loosely defined ideas. “Use AI for contract review”, “improve student support” or “help the marketing team” may sound promising, but they are not yet implementable use cases.
The ONS identifies difficulty finding appropriate business use cases and a lack of expertise as persistent barriers to adoption. Office for National Statistics
A useful AI initiative starts with a specific moment of work:
- Which task, question or service needs to improve?
- Who currently performs it?
- What information do they need?
- Where does unnecessary time or inconsistency arise?
- What would a better outcome look like?
- How will the organisation know that the change has worked?
This moves the conversation from what AI can do in theory to what should improve in practice.
The AI lacks organisational context
A general-purpose AI tool can be useful without knowing anything about the organisation using it. That is also its limitation.
It does not automatically understand the organisation’s policies, terminology, previous work, quality standards, client commitments or specialist expertise. Employees must provide that context themselves, often repeatedly and inconsistently.
For knowledge-intensive organisations, much of the potential value lies in connecting AI to trusted, approved information.
That could mean helping employees find answers across internal guidance, supporting lawyers as they work with established precedents, giving students access to institution-approved learning support or allowing clients to explore a curated body of professional knowledge.
The quality of the experience then depends on more than the underlying model. It depends on the quality, relevance and organisation of the knowledge surrounding it.
Experiments are not converted into shared capability
Employees frequently discover useful prompts and techniques, but those methods can remain on individual devices or inside personal chat histories.
When that happens, colleagues solve the same problem repeatedly. The organisation cannot test the approach consistently, improve it centrally or make it available to a wider audience.
Moving beyond experimentation means capturing what works and turning it into something repeatable: a configured chat assistant, a shared workspace, an approved prompt structure or a defined workflow for a particular type of task.
This does not eliminate individual judgement. It gives people a stronger and more consistent starting point.
Governance is treated as a final approval stage
Governance is often positioned as the point where a completed AI initiative is reviewed by legal, risk or information-security teams.
By then, important decisions may already have been made.
Responsible implementation should define from the beginning:
- Which information the AI can use
- Who should have access
- What the AI should and should not answer
- When users must check the original source
- Which outputs require professional review
- How feedback and problems will be recorded
- Who owns the experience after launch
Designed well, governance does not simply restrict use. It gives employees the confidence to use AI within clear, practical boundaries.
Activity is measured instead of value
Licences, logins and prompt volumes can indicate engagement, but they cannot show whether work has improved.
The right measures depend on the use case. They might include:
- Time saved on a recurring task
- Faster access to reliable information
- Improved consistency across outputs
- Fewer routine questions reaching specialist teams
- User satisfaction and repeat usage
- The quality of source citations
- The number of appropriate escalations
- Increased engagement with a client or employee service
Adoption matters, but adoption should be connected to an outcome.
The three levels of meaningful AI adoption
A useful way to assess progress is to consider three levels.
- Access: Employees can use an approved AI tool and understand the basic rules surrounding it. This creates a safe environment for learning and experimentation, but most value remains individual.
- Application: The organisation identifies specific use cases and configures AI around defined users, knowledge and outcomes. Examples might include an internal assistant grounded in HR policies, a workspace for document analysis or a structured workflow that produces consistent outputs across a collection of files. At this level, successful methods become repeatable and shareable.
- Transformation: AI becomes part of how the organisation delivers work, shares expertise or serves its clients. This could include giving employees faster access to specialist knowledge, redesigning a high-volume process, providing personalised support at scale or creating a new digital service based on the organisation’s expertise. Transformation does not require fully autonomous AI. A well-designed knowledge assistant or structured document workflow can materially change how work is delivered while keeping people firmly in control.
The important shift is from asking individuals to find uses for a tool to designing AI experiences around valuable organisational outcomes.
What leaders should do next
Organisations do not necessarily need more AI tools. They need to make better use of the technology they already have, or be clearer about what an additional platform would enable.
Leaders can begin by asking five questions:
- Where is valuable knowledge currently difficult to find or apply?
- Which recurring tasks involve avoidable searching, rewriting or manual processing?
- Which employee or client experience would benefit from faster, more personalised support?
- What information, boundaries and human judgement would that use case require?
- Which measurable outcome would justify expanding it?
Choose one bounded use case, test it with real users and approved information, and establish a baseline before launch. Collect feedback, assess output quality and improve the experience before extending it to more people.
This approach may appear less ambitious than a large organisation-wide rollout. In practice, it creates stronger evidence, clearer ownership and a more credible route to scale.
Turn one promising AI use case into measurable value
If your organisation already has access to AI but is struggling to move beyond individual experimentation, Kalisa can help you make the next step practical. Transform your trusted knowledge and proven ways of working into secure chat agents, shared workspaces and structured workflows that teams can use consistently, and that your organisation can govern, improve and measure.
Book a demo with our team to see how a real use case from your organisation could work on the platform and what it would take to move it from experiment to dependable, organisation-wide capability.
Powering the next generation of professional services
Kalisa offers everything you need to deliver valuable GenAI experiences to your clients and team.
- Knowledge agents with subject-matter expertise
- Sandbox Agents for day-to-day work
- AI Workflows to automate business processes
- AI workspaces for your team
- Self-serve client portals and dashboards
- Subscriptions and monetisation
- Analytics to measure usage and engagement
- Securely combine public and private data
- API for systems integration
%20(1).webp)
