Pilot Boldly, Commit Slowly

I was talking with a colleague the other day about AI and the risks we face. I said "we need to pilot boldly and commit slowly." I didn't realize until I'd said it how very apt it was (so yes, you can quote me).

In the very volatile world of technology and AI, there's risk on all sides.

The algorithms used today will continue to evolve quickly. Whatever AI is today, it will not be tomorrow. It will change a hundred or a thousand or a million times. This is the new Wild West. This is a common pattern we see in technology innovation, so let's take the lessons of the past and apply them to the situation today.

Risk in Committing

There's tremendous risk in committing to a particular AI path at this moment in time. Whatever you deploy will have issues. It will need to learn. It may contain hidden biases or errors. It may give wild answers from time to time. It will be influenced by the numbers and types of prompts. This is an emerging area of technology that is fascinating and in high demand, but for healthcare IT, it brings with it significant questions about potential risks.

Risk in Doing Nothing

There's also a serious risk in doing nothing. AI will continue to change and evolve at an accelerating pace and organizations that do not embrace it in some manner will find themselves being left behind. Healthcare IT rarely wants (or needs) to be on the bleeding edge, but there's risk is simply sitting back and waiting for all this to settle down.

Pilot Boldly, Commit Slowly

One of the ways to mitigate both risks is to embrace pilots in a safe environment. Whether that's piloting AI for automating certain repetitive operational functions or piloting more mature AI such as transcribing visit notes or drafting patient text responses, there are many different ways you can embrace AI today. It's great when we can say YES to our end users and help them discover how AI can potentially help drive better, faster, less expensive outcomes.

Set up clear parameters around a pilot. Specifically state that you're doing this to learn, not to implement. You'll set strong expectations about outcomes and you will less likely find yourself pressured into implementing a solution that is either not yet ready for primetime or simply just not what the organization needs.

In this "Wild West" atmosphere (or perhaps it's more like the Gold Rush period), it's easy to get swept up in the enthusiasm of potential. It's also easy to get a bit jaded when solutions don't live up to inflated expectations. Piloting is your path forward. No, you can't pilot everything, but piloting allows you to safely say YES to experimenting with these emerging solutions without committing. If a vendor is not willing (or able) to facilitate a pilot program, think twice. Nothing against the vendor, but it's a certainty that the solution you're looking at today will be vastly different in 12, 18, or 24 months. Consider that before you ink a multi-year contractual commitment.