This blog first appeared as Steve Wunker's piece for Forbes Generative AI like ChatGPT is truly exciting, and it’s easy to be seduced by the technology’s potential to produce, well, almost any sort of output. Be careful. The opportunity in generative AI is enormous but requires careful analysis of where the best applications lie. Healthcare, in particular, requires this assessment – this isn’t an industry known for fast change, and the risks of inappropriately deploying new technology can be huge. For instance, consider the hype around IBM’s Watson Health a few years ago; this AI was going to figure out complex cancers! It didn’t, and it was sold off cheaply in parts last year. With healthcare, we can deploy a straightforward, five-part approach to evaluate where enterprise-ready generative AI will gain traction early on: 1. Start with the problems the technology can help to address; what is it really good at doing? 2. Search for the big areas that have those problems 3. Understand the triggers and obstacles to adopting the technology in those top use cases. This includes what people need to stop doing in order to start embracing the new solution. (For Watson, oncologists were definitely not going to stop diagnosing cancers) 4. Assess the business dynamics around how the high priority categories will be entered 5. Look broadly at the levers for creating a full solution, including the technology but also going beyond to include, for example, workflow consulting, patient education, and much more To deploy this approach to healthcare, let’s be clear that we aren’t talking about other forms of deep learning, like the digital interpretation of medical images or assessment of population health datasets. Those uses are already well underway. Nor are we looking at simple applications of task-specific AI such as scheduling appointments. This article focuses on generative AI, and in healthcare that is quite new. Second, what big areas do these problems correspond to? For example, with those four domains we have examples like: a) Interpreting unstructured data: summarizing key facts in physician notes in an Electronic Health Record, asking health insurance companies for Prior Authorization of treatment, and seeking patterns in clinical trial data e.g. in Patient Reported Outcomes or among non-responders to therapy b) Explaining data in a coherent way: customer service for health insurers, providing diagnoses, and generating treatment plans c) Engaging people in conversation: obtaining screening data (e.g. do you feel safe at home?) and providing talk therapy for low-acuity behavioral health issues d) Generating new ideas: working with datasets on proteomics and genomics to discover new drugs and new ways to use existing therapies Third, what are the triggers and obstacles to adopting the new technology? This question will quickly reduce the likelihood of certain use cases in the near future. For example, until companies go the route of gaining FDA approval of generative AI as a medical device that can provide advice on specific courses of action, AI is not going to be providing definitive diagnoses or creating treatment plans for American patients. (There are many other obstacles, too, as Watson Health discovered). However, the outlook may be different in emerging markets where clinicians are overwhelmed by patient demand and regulatory requirements are less strict. The analysis can also point to areas that fit the pattern of rapid innovation adoption (few dependencies, high need, and low risk/switching cost). That would include self-paid talk therapy, for instance.
Fourth, what do business dynamics indicate about which markets will be entered first? The complexities of that question aren’t well suited to an article-length piece, but the answer would comprise factors such as unit and scale economics, channels to market, the sales process, and competitive intensity. Finally, think about the full solution. Rarely is a truly novel technology like generative AI enough to get people to change long-running work practices. Significant uptake can require training customers and building an ecosystem of complementary offerings, for instance. Moreover, that encompassing solution also helps companies to differentiate their offerings when rivals inevitably mimic the underlying technology. If you work in a healthcare or life sciences enterprise and can focus on particular contexts, you may also take a different approach. Rather than starting with the technology, focus on key challenges and look holistically at what type of solution is really required. Then you can examine ways to achieve those aims; generative AI may be one, but there could be other lower-tech approaches too. In healthcare, the near-term opportunities for enterprise-grade generative AI abound. This five-part approach illustrates how diverse the use cases are. Even in such a conservative industry, we can expect transformative change, and soon. By Steve Wunker Comments are closed.
|
3/9/2023