Generative models have been around for a while, but its mainstream adoption is what will drive these transformations at scale.
Its applicability across industries already is visible, so healthcare as an industry is no different. Be it in improving the patient experience by introducing conversational appointment-setting processes, or using summarizations to enable doctors to review the appropriate information at the right time, or improving the overall efficiency in the hospital care system, each of these implementations are having profound effects on the way we think about patients and overall healthcare.
While it is early days and there are lots of things still to unfold – such as which use cases have the least hallucinations [errors] and what data privacy and sharing rules we will need to define in order to ensure the ethical nature of the models, among other things – the results are promising.
In terms of whether it can be reliably used for every use case one can think of, probably not and certainly not yet. But is there a path to get us to that stage, with the right governance and controls, certainly yes.
challenges generative AI faces to getting to mainstream use?
. There are several:
- Lack of clarity in where to apply. Across businesses, urgency to start adopting gen AI is making them pick nonvalue-adding use cases. For example, time and again we are seeing conversations like, “We need to do something with gen AI, our board wants it.” And when asked exactly what, in many instances there is no clear answer. Another example is we will say to clients, “In order to do ‘X’ we need to get your data in order,” and customers say, “Well, that is not a priority. We cannot do that.” So, there are several disconnects that need to be addressed.
- Hallucinations. Most of these models are hallucinating. There is a good bit of work required to get them to a stable state for relevant business functions. Most businesses may not have the appetite for a slower rollout to get these wrinkles ironed out.
- Lesser line of sight to ROI. Currently consuming most of these models is expensive and the cost of implementing is high due to talent shortages. Companies are struggling to be very clear on how a large upfront cost can be justified for a longer-term ROI. Until economies of scale are reached, access will be limited to organizations that are able to carve out budgets, thus limiting mainstream adoption.
What should executives and clinicians at hospitals and health systems do when faced with new generative AI tools?
. Establish gen AI checks into your vendor onboarding. Evaluate your vendors’ security, compliance and governance policies thoroughly. It almost requires a rewrite of internal guidelines to accommodate all the nuances coming with AI technologies. Also:
- Have zero trust policy. Stricter compliance audits around what is being shared across IT providers, how is the visibility provided back to you as well as how the data is internally processed.
- Invest in internal experts. The space is new and everyone is learning as it evolves. Investing in internal experts and partnering with the right industry partners who can help you navigate the evaluation and deployment processes across these vendors will be important in the success of these launches.
Oracle Clinical Digital Assistant generative AI services as an example
Oracle is introducing a set of generative AI services for healthcare organizations within its EHR solutions portfolio. Particularly the vision that is laid out for the Oracle Clinical Digital Assistant in how it will aim to streamline administrative tasks with voice commands will certainly be revolutionary in how you will experience clinical visits in the future.
While the vision is great, we will need to see the practical rollout of the features that are going to be around 8-12 months out. I firmly believe though that the investment is in the right areas. If you are visiting a doctor, you want the doctor to be significantly prepared ahead of your visit, ensure that they spend less time looking at the reports, etc.
In addition, the patients themselves should have the ability to get as much information they can about their conditions, causes, etc., as easily as possible. Having such a digital assistant not only frees up time for administrators and clinicians but also helps maintain a seamless experience for the patients.
Where generative AI in healthcare would be five years from now?
We would certainly see revamped business processes where gen AI is mainstream. A lot of newer solution offerings and technology players will emerge that will dominate the market in very niche areas.
We will see a complete overhaul of the compliance and regulatory guidelines to monitor bad actors as well as protecting individual rights.
Most importantly, the world will move toward a lot of preventive care helping us proactively monitor our wellbeing, which in turn will improve the life expectancy of every individual.