Forget the Hype, What do Doctors Really Want from AI?
Everyone is talking about AI disrupting healthcare. But what are the real issues clinicians need solving? Jia Lin Yong spoke to eight clinicians across the UK to hear what they really need from new companies…
The previous digital health wave moved large portions of healthcare online. Today almost 30% of health appointments take place virtually in the US and the UK, with many health tech unicorns emerging by seizing this opportunity. Now, there is a new wave. The proliferation of large language models, the wealth of under-utilised health data, combined with a clinician shortage and increased demand for healthcare, creates a revised “why now” for entrepreneurs to build in healthcare. In particular health data, which is not publicly available, holds real and untapped value, such as predictive analytics and personalised insights, that big tech have a lower likelihood of dominating.
AI can, and has already created breakthroughs in healthcare, but so often new solutions in this space are built without clinicians - the users - in mind. Clinicians’ voices are underrepresented at health tech conferences, blogs, and podcasts about AI and health. This results in a mismatch between the problems needing urgent solutions and the technology being created, and can lead to adoption challenges for startups as they try to ‘build things people want’.
In order to better understand the solutions that needed to be built and funded, I interviewed eight leading clinicians, listed at the end of this article, to gather their insights. I asked them questions about the problems that need solving, health tech solutions that have, or have not worked, and their advice for founders. Here’s what I found.
Workflows and systems today are not fit-for-purpose
Almost all clinicians I spoke to mentioned that the biggest problem they face day-to-day is not having the right patient with the right information at the right place, at the right time. Clinicians spend much of their time chasing information from others, which is worsened by the lack of effective communication systems. During a hospital visit, I learnt that large amounts of communication is still done on Whatsapp (using personal numbers) or even on paper (e.g. I saw a ward change noted on a piece of paper).
Fragmented data and disconnected systems create significant inefficiencies in clinical practice. “It makes it hard to deliver whole-person care as without all the information we need in front of us, we end up looking at things in isolation”, one clinician told me.
These outdated systems have a real impact on patient care. One clinician mentioned that older systems mean that it can take 15 minutes to administer a simple prescription, and another mentioned that they have to log in to six separate systems before starting their day. All this is after assuming that “the computer switches on in the morning, my login works, and the internet stays on.” Multiply this by thousands of doctors and millions of patients, this quickly becomes a critical source of (sometimes direct) patient harm and wasted financial resources.
Clinicians are spending less time doing what they love most - treating patients
The administrative burden on clinicians has grown considerably - clinicians in the UK spend on average, a third of their time on admin, whilst in the US, this figure is closer to 49%. Interestingly, clinicians shared that the introduction of electronic health records (EHRs) and digital forms has in fact exacerbated the administrative strain. For example, a well-intentioned digitised asthma referral protocol that aimed to reduce manual work has added more burden and bottlenecks. After all, digitising forms will not save time if it simultaneously introduces more log ons, more clicks, more data entry, and new processes.
One clinician shared that they are in a constant battle of balancing on-the-day care, patients wanting continuity, and time spent processing letters, prescriptions, and other forms of admin. This corroborates with junior clinicians I met who have left the profession, disillusioned with medical practice being mostly an administrative job. “We spend far too much time looking into or typing into a screen”.
Another essential part of clinicians’ roles is building a relationship and trust with patients. Patients are often going through an emotionally challenging time. All clinicians mention the complexity of dealing with the emotional and logistical concerns of patients and their families - on top of day to day workload. “One day I found my patient’s family member feeling very distressed, and it was because they couldn’t figure out how to pay for the parking”.
In understaffed hospitals and clinics, it is challenging for clinicians to do it all. Over time, clinicians are spending less and less time doing what they love most - treating patients. It is no wonder clinician burnout is on the rise.
Clinicians are excited about how technology will help them
It is a common misconception that clinicians are resistant to change and unwilling to adopt new technologies. Clinicians are ready to embrace new technologies and are enthusiastic about how they can help ease their workload and free up their time for patient care. In fact, “once clinicians are convinced about a new tool, they become its biggest advocate to other clinicians as well as budget holders”, one clinician told me We discussed some considerations for founders to make when starting AI x healthcare companies, and there were some clear recommendations:
1/ Solutions need to address a core need
Too often clinicians are pitched tools that do not address real pain points. “If it is not in their top 1-2 priorities, they will tell you that they like it but they won’t buy it”. The clinicians I spoke to encouraged founders to shadow clinicians to better understand their pain points, and work with them to develop and iterate their products. Accurx’s text-based system was cited by almost all clinicians as a solution that got it right. It is very simple but fixes a problem that clinicians really wanted to be solved.
2/ Minimise complexity
New tools and systems should be designed for ease of use and to minimise cognitive load for clinicians. Founders should be cautious of the “additive bias”, which is the tendency to believe that adding features is better than removing them. One clinician urges founders to adopt a systems-thinking approach when identifying root-causes of inefficiencies and developing new products. Sometimes the answer is in simplifying and redesigning a system from first-principles rather than adding new features. For example, Elea (Giant portfolio company) is systematically designing AI-native pathology workflows to bring down time to diagnosis from weeks to days. The Concentric app is another example provided by clinicians as an uncomplicated solution that has resulted in significant improvements in efficiency. It simply streamlines patient consent processes to reduce paperwork and reduce the likelihood of lost/incomplete forms.
3/ Easily integrates with existing workflows
New tools and systems need to seamlessly fit into clinicians’ daily routines and workflows to maximise adoption. In recent years, the introduction of ‘add on’ tools that are not interoperable means that clinicians sometimes have to waste time copy-pasting information from one system to another. “There is a belief that if there is a tool, people will use it, but if it doesn’t integrate quickly, easily, and talk to other systems, it won’t get used”. New tools should almost feel invisible. For example, one clinician mentioned that ambient pervasive AI documentation assistants that accurately captures all necessary information during patient interactions and seamless integrates into the electronic health record could save hours of work and potentially improve the quality
4/ Concrete return on investment (ROI)
A tool that saves clinicians five seconds but requires days of training and thousands of pounds spent on change management is unjustifiable. There is also an assumption that AI is always faster than humans. However, after 15-20 years in practice, clinicians develop an intuition that makes them faster than most machines. Many AI and automation tools have also failed at the clinical safety hurdle as the time and cost associated with the additional checks and balances required, often outweighs the benefit of the tool. It is important to not look at the value delivered to the customer in isolation of other system changes required e.g. IT team implementation, clinical safety and governance systems, etc.
5/ Distribution is King
Often it isn’t the startups with the best products that win, it is those with the best unfair distribution advantage. It is vital for founders to start thinking about distribution early on in their journey. In the UK, for example, the decision-making structures and procurement processes are often challenging to navigate. It is not only the sales process that is opaque, but it is hard to build a landscape view of tools adopted today, for example, to avoid duplication or understand integrations required.
Founding teams with strong networks with decision-makers may find it easier to get in front of the right people. However, many founders will likely need to be creative with distribution and partner with other providers (channel partnerships / joint development) in order to secure reach to market. Some startups focus on grassroots adoption and offer tools to clinicians for free and only charge after mass adoption. The benefit of this approach is that users are generally very sticky and startups can gather more ROI and/or engagement data to accelerate sales. Lastly, whilst most startups typically focus on getting buy in from one or two key opinion leaders (KOLs) or well-known professors, some clinicians suggest that they may not always be the right people to go after, and that trainee doctors who are tech native and faster to adopt may be better champions of new tools.
6/ Don’t forget the patient
One of the unique challenges of building in healthcare is that the problems being solved are multi-stakeholder, each with slightly different priorities. Whilst most of the discussion was focused on clinician-centric design, many clinicians flag the importance of understanding the value to patients.
Now is the time to build in healthcare. Health systems are accelerating their adoption of new technologies, AI capabilities are advancing at an exponential rate, and the opportunity that can be captured is significant (dare we believe that the next $100bn+ company will be built in healthcare?). However, the right guardrails need to be put in place. Building in healthcare can be rewarding - both from an impact and financial perspective - however, it can only be achieved with a thoughtful approach to building that takes into account the various stakeholders, from payers, providers, patients, and importantly, clinicians.
Special thanks to the following clinicians for their generous insights:
Dr. Matthieu Komorowski, Intensive Care and Anaesthesia Consultant at Imperial College Healthcare NHS Trust, Senior Lecturer at Imperial College London
Dr. Osman Bhatti, General Practitioner and Chief Clinical Information Officer at NHS North East London ICB; GP VTS Programme Director at NHS England
Dr. Johnson D’Souza, PCN Clinical Director, General Practitioner at Valentine Health Partnership
Dr. Umang Patel, Chief Clinical Information Officer at Microsoft, Paediatrician at Frimley Health NHS Foundation Trust
Dr. Shanker Vijayadeva, GP Lead for Digital Transformation at NHS England (London region), GP Lead for Ealing Borough and North West London ICB
Dr Nikita Kanani MBE, Chief Strategy & Innovation Officer at Aneira Health, General Practitioner at St. John Medical Centre
Dr. Keith Grimes, Founder of Curistica, former NHS General Practitioner
Dr. Nicola Turner, General Practitioner Partner at The Hall Practice, Board Director at FedBucks