Reading the NHS 10 year plan I have to applaud its vision and the drive for a digital doctor in everyone’s pocket. There is no disputing that it’s a good idea.
However, what I can’t see in the 10 year plan is how the government intends to make sure that the digital doctor in your pocket is equally accessible to everyone. It doesn’t explain how it will make sure that everybody has a smart phone in their pocket that will be able to download and use the digital doctor.
Digital poverty is a big issue, with over nine million people having limited or no access to digital in England – this being defined as ‘the inability to fully interact with the digital world, when, where and how needed.’
Whilst this isn’t stated in the plan, I have heard NHS and DHSC leaders suggest that one of the roles of charities will be to make sure that people who don’t have a smart phone, will be able to acquire one via a local charity. This seems a bit unrealistic, given how much bigger the NHS budget is vis-a-vis the budgets that local charities have available. Also, who will pay the monthly network provider costs? Or will network providers be expected to provide ‘free access’ by putting up charges for everyone else?
There seems to be some recognition by the NHS (but not in the plan) of the need to ensure that digital infrastructure such as mobile networks and broadband networks are accessible everywhere and that people who currently live in digital dead zones will have coverage. However, this seems to be something that’s left for local organisations to solve rather than being a national priority.
Getting every NHS organisation to use the same single unified digital patient record system will be a challenge. There are four or five main systems that seem to be in use and that is without considering all the different digital care records used by local government and social care as well.
Which NHS trusts are going to have to give up the digital patient systems that they have spent years investing money and staff training time on? How and where will they be able to find the money to make these changes? These things are unclear.
I was recently involved in some work with a number of councils looking at how we could improve advice and guidance for local people. Most of these councils were keen to move to a digital only approach. What we found from listening to local people was that this wasn’t people wanted. Whilst people valued the ability to be able to access digital advice and guidance, they also wanted access to real people as well.
This isn’t to say that the digital doctor it isn’t a good idea. However, it needs to go alongside a vision and a policy drive to get rid of digital poverty and to improve digital literacy for everyone. I’m not only talking about older people who haven’t embraced the Internet. I’m talking about people in many walks of life who can’t afford a smart phone, or the cost of running a smart phone, or of having Wi-Fi at home.
When we look at the use of AI as a way of improving productivity and reducing the number of clinicians needed (which seems to be part of the plan’s rationale for embracing AI) or giving clinicians more time to spend with patients; there are some examples out there which illustrate how this may not happen in practice.
Here are some of the companies that implemented AI or automation strategies but later reversed course and reintroduced or expanded human roles—usually due to issues with quality, customer satisfaction, or limitations in the technology.
Amazon. Amazon invested in warehouse automation and AI, particularly with its acquisition of Kiva Systems for robotics. Despite the expansion of robotics, Amazon hired hundreds of thousands of warehouse workers and has publicly stated that humans will remain central. Amazons CEO admitted that robots still cannot handle the complexity of many warehouse tasks. Robots struggled with tasks requiring fine motor skills and complex decision-making. Human workers remain essential for flexibility and adaptability.
IBM Watson Health. IBM launched Watson Health with bold claims about AI transforming healthcare and even diagnosing cancer. IBM eventually sold off Watson Health in 2022, after hospitals and clinicians found the system underdelivered on promises. The AI lacked accuracy and contextual understanding in clinical environments. Human doctors were still needed for nuanced, evidence-based decision-making.
Zappos. Owned by Amazon, experimented with AI-powered chatbots for customer service. The company recommitted to human-led service which had been a hallmark of its brand. Customer feedback indicated frustration with bots. Human interaction led to higher satisfaction and loyalty.
Meta (Facebook) Meta heavily invested in AI to moderate harmful content on Facebook and Instagram. Despite continued AI use, Meta had to hire thousands of human moderators. AI failed to detect nuance in hate speech, misinformation, and cultural context. Human reviewers were more effective for borderline or complex content.
Air Canada. Used an AI-powered chatbot on its website for customer service queries. In 2024 a Canadian tribunal ruled Air Canada was liable for incorrect information that the chatbot gave to a customer—despite the airline claiming the bot “made a mistake.” This incident highlighted the risks of relying on AI without human oversight, prompting reevaluations of chatbot autonomy across sectors.
McDonald’s. Tested voice-ordering AI systems in drive-thrus in partnership with IBM. In 2024, McDonald’s pulled the plug on AI ordering, in over 100 test locations in the U.S. The AI had trouble understanding accents, complex orders and background noise, leading to order errors and customer frustration.
Cleveland Clinic. Adopted AI to help read radiology scans. While still using AI, the hospital emphasised reintroducing more radiologists for oversight. AI was helpful as a tool but not reliable as a sole reader of scans. Human expertise remained crucial for interpretation and patient care.
The learning from each of these examples is
Quality and Accuracy Issues. AI under performs in real-world conditions (e.g. Amazon’s warehouse robots, McDonald’s voice AI, Watson Health’s diagnoses), machines struggled with complexity, ambiguity, and edge cases. Humans were needed for tasks requiring nuance, contextual understanding, or fine motor skills.
Customer Experience Problems. Customers often preferred human interaction, especially in service and support roles (e.g. Zappos, Air Canada, McDonald’s). AI interfaces, such as chatbots or voice systems, frustrated users when they couldn’t resolve issues or understand context.
Legal and Accountability Risks. Companies like Air Canada learned that AI-generated mistakes can result in legal challenges. Firms reassessed the risk exposure of letting AI act autonomously without clear human oversight or accountability structures.
Inability to Replace Human Creativity or Empathy. In sectors like healthcare, education, or content moderation, AI lacked the empathy, ethical reasoning, or emotional intelligence that humans bring to sensitive or judgment-based tasks. Meta’s moderation problems and IBM Watson’s limitations in clinical settings are key examples.
Brand and Trust Considerations. Some companies, like Zappos, found that AI conflicted with their brand identity, which emphasised human-centred, personalised service. Over reliance on automation risked damaging customer loyalty and trust.
Limitations of Current AI Technology. AI systems often struggled with variability, especially in noisy or unpredictable environments (e.g. drive-thrus or warehouses). Current AI models are often brittle outside their narrow, well-defined domains, leading to reintroduction of humans as fallback.
Operational Flexibility. Human workers often proved more adaptable than AI systems when tasks changed or needed on-the-spot adjustments. In warehousing and logistics (e.g. Amazon), this adaptability was essential during peak periods or supply chain disruptions.
Reports over the years have highlighted that the NHS has spent billions on IT systems, some of which have been criticised for not delivering the expected benefits. For example, the National Programme for IT (NPfIT), which was launched in the early 2000s was ultimately deemed a failure, costing approximately £10 billion.
I am all for using digital technology and AI to improve health care and social care. However, I don’t think that digital technology and AI will lead to the NHS or social care needing a smaller workforce. It may even lead to both services needing to expand their workforce. We also need to actively focus on digital poverty. Could one of the answers be digital technology on prescription?
Jim Thomas
July 2025
Leave a Reply