In early June, the UK government announced a £1.1bn ($1.47bn) push to strengthen Britain’s AI computing power, funding a new national supercomputer and backing domestic semiconductor firms. It is the clearest indicator yet that AI is a national priority, and the message to business is clear: adopt AI or be outpaced by your rivals.
What I am hearing from business leaders is that the discussion around AI has taken a new turn. Companies now need to think beyond AI as just a productivity tool and consider what kind of enterprises the technology will create, and whether they are prepared to redesign their organisations around it.
I think many businesses still approach AI as a patchwork collection of tools such as copilots for employees, assistants for workflows, pilots in isolated teams and third-party applications layered onto existing processes. These projects can deliver temporary improvements, but they rarely change how the enterprise fundamentally operates.
I think the real opportunity is much bigger, notably the chance to build organisations in which human decision-making and machine execution work as one system, taking companies to new levels of productivity and profitability.
I have already seen this happen in healthcare, where automated test-scenario creation and data generation have supported platforms serving more than one million dual-eligible members across 25 states. This was achieved by reducing low-value execution and improving speed and consistency.
It is happening in software delivery too, with AI taking over much of the heavy lifting in repetitive tasks such as validating code changes, documenting updates, and moving information between development tools.
I think healthcare and software development are just the beginning, though. Other industries are likely to achieve similarly striking results soon, but first, there are some hurdles to overcome.
The core bottleneck
When AI programmes stall, organisations often blame the technology, citing LLM hallucinations, inaccurate outputs or uncertain returns. While some of those concerns are valid, they are rarely the main reason value fails to materialise.
The problem I see most often is not model quality, but operational ambiguity. Companies introduce AI into environments with unclear ownership, fragmented workflows and no shared model for accountability. In those conditions, even capable systems find it difficult to generate lasting value.
Many leadership teams also mistake activity for progress. They track the number of tools launched, prompts written or use cases tested, without asking whether work is being redesigned, decisions are being clarified, or responsibility is becoming more explicit.
From AI tools to AI operations
This shift from perceiving AI solely as technology that powers software to seeing it as a new operating layer inside work, will be accelerated by the adoption of agentic AI. These are systems that take AI way beyond just generating content or answering questions into orchestrating tasks, moving information across systems, supporting decisions and executing repeatable workflows with speed and consistency.
While this creates a new opportunity for the enterprise to rethink its approach, it also raises important new issues around the relationship between humans and machines. I believe that in the future, humans will remain responsible for key criteria such as direction, judgment, exceptions, ethics, and accountability, leaving more mundane tasks to machines.
Over the past decade, companies have digitised existing workflows without fundamentally rethinking them. AI creates a different opportunity in that it permits leaders to revisit how work is sequenced, how handoffs happen, which decisions should be automated, and where human involvement genuinely adds value.
Scale without losing control
The moment AI moves from experimentation to core workflows, concerns become more serious, as companies need to consider trust issues and operational risk. For example, operating agentic AI systems in highly regulated industries, such as finance or healthcare, can create traceability issues with serious consequences. I do think, too, that embedding agents in unstructured or unsuitable ways within workflows can create confusion or, worse, have a detrimental impact on staff morale.
For some business leaders, the issues seem so overwhelming that they choose not to engage and instead postpone AI implementation, pencilling it in for a later date.
Business leaders in other organisations attempt to engage with the process but mistakenly treat trust and scale as conflicting forces. I don’t see it that way, in fact, I believe trust is what makes scale possible.
If no one knows who owns the output, how a decision was reached, or when a human needs to step in, the system will not be trusted and will not scale. On the other hand, if accountability, openness and oversight are built into workflows from the start, AI becomes much easier to operationalise.
Redesign workflows, not just interfaces
Another reason so many organisations struggle to create value from AI is that they expect it to transform old, and sometimes archaic, workflows.
That rarely works, and if a workflow is fragmented, slow or dependent on too many handoffs, adding AI may improve parts of it, but it will not address the inefficiencies intrinsic in the system. Real gains come when leaders redesign the flow of work itself by reducing unnecessary steps, assigning repeatable actions to agents, creating clearer escalation routes and freeing people to focus on decisions that require experience, context and creativity.
There are many places where an agent-operated model can start to make a measurable difference, most importantly by limiting the time and energy that valuable human capacity is consumed by low-value tasks.
Make AI a management practice
Treating AI seriously means embedding it into management practice. AI has to become part of how the wider business is managed. Leaving it siloed in technical terms will prevent its adoption and implementation.
It also requires a new approach to workforce development. Employees need ongoing exposure to AI in real workflows, with practical guidance on when to trust it and support in building judgment as their roles evolve.
They need to understand how to work with AI responsibly, know when to challenge it and how their own contribution changes as agents take on more execution. And as AI evolves, they need to be constantly learning about the new opportunities it creates and how to maximise them.
In our own experience, this is where many companies still underestimate the task. It is relatively easy to buy access to AI. Creating an organisation in which thousands of people can use it consistently, confidently and productively is a serious leadership challenge.
Why leadership decides the outcome
In the end, recalibrating businesses to achieve optimum performance from both humans and machines comes down to leadership. Enterprises will be defined by who took responsibility for AI and turned it into a coherent operating model.
Leaders must decide where AI matters most, set the guardrails for its use, align incentives across functions and ensure the organisation builds the capabilities to use it well.
The companies that pull ahead will be those whose leaders treat AI as a redesign of how their enterprises operate.
AI will continue to evolve, and the hype cycle surrounding it will keep moving. The underlying shift will not.
The question for leaders now is whether they are still experimenting with AI at the edges or ready to reorganise the enterprise around it.
