The surgery that made GPs a specialist resource and the AI implications

What's the GP in your business

Healthcare has been one of my focus areas for the last five years. One model has stayed with me more than any other: Cuckoo Lane Practice in Hanwell, West London.

It is nurse-led. Nurse practitioners deliver much of the day-to-day care, supported by nurses and other clinical staff. A GP is available for several hours each day for cases that need specialist input. NHS England has documented the model, and the practice has received an Outstanding CQC rating. NHS England, CQC


  • The point is not that GPs are unnecessary. Far from it. The point is that their expertise does not need to sit at the front of every interaction.

  • The model allocates scarce expertise to the cases where it adds most value.

  • Now add AI.

AI can support assessment, triage and clinical decision-making in defined settings, provided it is validated, governed and used with clear clinical accountability. It does not replace professional judgement. It can, however, help capable clinicians work with better information and direct complex cases to specialists sooner.

That leads to a question I now put to leadership teams:


What is the GP in your business?

Every organisation has one.

It may be a role, process, system or approval stage that everyone assumes is essential. It often sits at the centre of how work gets done. AI may not remove the skill itself, but it may make its current place in the operating model optional.

Find it, and you may find the real value of AI. Miss it, and you may spend the next five years automating around it.

Find it, and you are on the road to AdvantageAI and real ebitda benefits!


Why organisations struggle to see it

Most AI programmes begin with technology.

They collect use cases, rank them, automate the obvious tasks, roll out copilots and run pilots. Each initiative may make an existing activity faster. Few ask whether the activity, hand-off or role should exist in its present form at all.

We have seen this pattern before.

Business-process re-engineering promised to rethink work from first principles. Yet many organisations used new systems to automate existing workflows rather than redesign them. AI creates the same temptation, only at greater speed.

The largest gains often come from removing stages of work, changing decision rights or reconfiguring how expertise reaches the customer.

That requires leaders to challenge work that people have built careers around.


The resistance is often the signal

I saw this clearly while working with a large private healthcare provider in Eastern Europe.

The group had already applied AI successfully in administration, onboarding and subscriptions. Those improvements mattered, but they did not change the core economics of care.

  • The more consequential conversation began with clinical work.

  • We worked with the CEO and leadership team to examine how AI-supported decision-making could improve the way care was delivered. The discussion was difficult. Doctors understandably worried about professional autonomy, quality and responsibility.

  • That resistance was informative. When the people closest to a process push back most strongly, they may be protecting the part of the operating model that AI will force the organisation to rethink.

The CEO’s role was not to tell clinicians that technology knew better. It was to set a clear expectation: AI-supported practice, properly governed and clinically accountable, would become part of how the group delivered high-quality care.

That is a strategic decision. It cannot be delegated to an IT project.


Every sector has a GP

The same pattern appears outside healthcare.

At an architecture and design practice, the traditional value proposition rested partly on carefully developing a small number of design options. AI changes the cost of exploration. When a team can test hundreds of credible options, the scarce asset is no longer the ability to produce options.


  • The question becomes: what do clients really value?

For that firm, AI created pressure on some traditional activities. It also created an opportunity to move further into delivery, helping clients ensure that the design they commissioned becomes the asset that gets built.


  • At a grocery business, the assumed constraint was a major core-system replacement. The accepted view was that change had to wait for a large technology programme.

  • Instead, the business used data and AI to work around parts of its legacy estate. It improved customer propositions and commercial decisions while avoiding years of delay.

In both cases, the “GP” was something everyone had accepted as fixed:


  • the expert who must touch every decision

  • the established output customers supposedly pay for

  • the system that must be replaced before improvement can start

None was as fixed as it appeared.


Why this cannot sit only with the CIO / CTO

AI is often assigned to the CIO / CTO or digital team. That is understandable. AI involves models, data, integration, security and technology choices.

But these teams cannot decide how the business should compete, what it should stop doing or which role needs redesigning.

McKinsey’s 2026 survey captures the gap. Eighty percent of respondents say AI has improved individual productivity, but only 37% report a positive contribution to enterprise EBIT. McKinsey

The time saved at the desk often disappears into an unchanged organisation.

At Veles, we use a simple rule of thumb:

  • 10% is the AI itself: models, tools and agents.

  • 20% is data and integration.

  • 70% is reinventing how work gets done: roles, decision rights, workflows and operating structure.

The CIO can lead much of the first 30%.

The CEO, supported by the CFO and leadership team, must lead the other 70%.


How to find your GP

You will not find it in a conventional use-case workshop. You find it by asking questions that make people slightly uncomfortable.

  1. What would a well-funded new entrant build today, with AI and none of our legacy?

  2. Which roles exist mainly because information used to be scarce, slow or expensive to interpret?

  3. Where is expertise rationed through queues, sign-offs, referrals or senior review?

  4. Which outputs have become dramatically cheaper to create?

  5. What do customers truly value, and which parts of our current model do they simply tolerate?

  6. Where is resistance to AI strongest, and what is that resistance protecting?

Once you have candidates, use three actions:

  • Kill work that exists only because the technology was not previously available.

  • Rebuild the delivery model around the customer outcome, as if you were starting today.

  • Invent services or capabilities that were not previously possible.

Then choose one change in core work and prove it in the P&L.

Not a platform. Not a transformation programme. One measurable redesign.


Start with an honest conversation

Every business has a GP.

It may be the senior approver who holds up decisions. The specialist team customers cannot access quickly enough. The report that takes weeks to produce. The legacy system everyone assumes must be replaced first.

You already hold advantages that a new entrant does not: customer knowledge, industry expertise, operational data and trusted relationships.

The question is whether you are willing to examine the parts of the organisation that those advantages have helped create.

What is the GP in your business? This is the key to finding the EBITDA from AI.

Eddie Short

View on LinkedIn