Alex Fairweather explores the new potential of applying AI to clinical practice and explores the difficulties of storing and using knowledge.
Knowledge management refers to the collection, storage and application of knowledge in explicit codified form or tacit personalised form. In healthcare settings, this may come from data, training, journals, research and clinical experience. Morten Hansen and his colleagues have argued in the Harvard Business Review that which method a firm chooses to manage its knowledge is key to its competitive strategy, drawing examples of different consulting firms using both forms successfully and also drawing parallels to healthcare as deriving competitive advantage in knowledge management.
Two types of knowledge are discussed in this context: codified knowledge which is stored and re-used, and tacit knowledge which is analytical and personalised.
In the example of consulting firms, both knowledge types involve very different IT systems, HR and economic model requirements. Healthcare is in most cases a combination of both types of knowledge this presents difficulties in productively applying technology to apply knowledge in practice.
How AI facilitates knowledge management
In the application of AI in healthcare, machine learning could be considered as codified knowledge management in arenas such as imaging and diagnostics; where large volumes of data are analysed. Generative AI, however, could be used also in personalised, tacit forms which confirms findings in this study that Gen AI could be most successfully used to augment human performance; human-to-human interaction being an essential part of healthcare and indeed a bottleneck that could be reduced through Gen AI application.
Knowledge management strategy is also important in considering either individual or organisational use of Gen AI; if individuals have access to web-based applications for example, far less IT infrastructure is needed to support this. If knowledge is codified and used more centrally then an advanced and powerful IT infrastructure is required to cope with the energy-intensive demands of Gen AI. Training necessary to use Gen AI applications could also be similarly affected by this choice – personalised, individual use can likely support a leaner training system, which, depending on the organisation, may lend itself to faster and more effective use.
The importance of data
Data is essential to train and refine AI algorithms and raise capital for tech startups; the availability of proprietary data has been linked with greater levels of VC funding, with data itself being a source of competitive advantage in many applications. Comparing data with accumulated experience and knowledge in a healthcare setting suggests individual AI applications built on this knowledge could also create a competitive advantage for practices. Though knowledge is in high supply from clinicians, codifying this and storing it in useful digital formats is challenging.
Tacit knowledge is by definition not stored outside of individuals or groups and accumulated data in healthcare settings is often fragmented, siloed and messy. As Yuji Roh and his colleagues have explained, data collection, preparation and labelling can be bottlenecks in the development of machine learning applications with data management techniques also contributing to problems.
Despite recognised issues with data, success cases have been recognised within healthcare centres’ access to proprietary data in collaboration with technology providers in developing tailored applications. In an article for California Management Review, Jürgen Brock and Florian von Wangenheim observed that digital leaders were strongly more advanced in seven organisational traits: “integrated data management, CEO priority, security strategy, digital processes, digital strategy, agility, and open innovation ecosystem”. The most important of those is integrated data management.
For healthcare organisations that intend to innovate with AI there must be internal capabilities to prepare and manage proprietary data or, possibly more realistically, the organisation must collaborate with external technology developers to bring in these capabilities and work on a defined application. The organisation must have suitable teams, leadership, subject matter experts and internal processes to capitalise.
Applying GenAI to private practice
This model of knowledge management suggests two possible applications to clinicians in practice.
First, using an existing LLM trained on large volumes of medical data may help clinicians apply tacit, personalised knowledge to a larger overall base of codified knowledge. If this can be done in the context of patient consultation this has great power to improve speed and quality of personalised care.
Second, codifying and storing accumulated knowledge gives practices and clinicians new possibilities to create a competitive advantage in their practice. Furthermore, this may give new capabilities to the longevity of owner-managed practices after the retirement of the founders or practising clinicians.
As AI becomes more widely adopted and access to it becomes more commoditised, it’s important that practices of all sizes consider applying at least some level of technology to maintain a competitive edge. As discussed previously, in the context of healthcare, the human element should never be ignored and new technologies should be considered to augment and improve the performance of human-to-human care. We are seeing new examples now weekly of how this can become more real and with good guidance can be diffused across the industry.
