Healthcare AI Insights from the SHI Summit

As I mentioned in my roundup of healthcare cybersecurity insights from the SHI Summit, I had the pleasure of mixing and mingling with the great SHI International community as they shared some great content related to cybersecurity and AI. Below you’ll find some of the AI insights that I found interesting and useful for the healthcare community. Plus, I added a little healthcare specific commentary.

It’s not very often that you see a privately held company the size of SHI International run by a female CEO, so it was pretty special to see Thai Lee on stage sharing SHI International’s efforts. Kudos to her and the team for all they’ve achieved and how they’ve embraced AI.

This stat was something we heard a few times at the conference. It’s kind of sad to hear that so many AI projects aren’t producing the outcomes that businesses want. As I mentioned in the above, I think healthcare does a bit better than this. Possibly because we’re a little more risk averse in how we implement AI. We’re thoughtful in how we’re implementing it and most healthcare organizations are focused on a clear outcome. They’re not just implementing AI to implement AI. They have a pre-defined outcome they’re looking to achieve.

I think that the speed of AI scares many in healthcare. What’s interesting AI right now isn’t that the transformation like what AI is providing is new. It’s the speed at which AI is transforming things. We’ve all been through transformations before, but AI is moving so much quicker than those transformations. That does present a unique opportunity if we don’t constrain it.

The popular phrase is that bad data in leads to bad data out. I like the example of how we’re training the data. The challenge is that there’s a lot of nuance in the data and we may not realize that the AI is missing the nuance in the training data we’re using. That’s a big problem if we’re not careful.

This is so perfectly stated. Everyone would love just one AI platform to easily implement in their organization and be done. They want the easy button to get the value out of AI. We may get there one day (I predict one AI platform per area of a hospital), but as we are today there’s definitely no easy button. There are a lot of buttons and not all of them produce the results we are hoping to achieve.

Isaac Blake had so many great one liners AI. What do you think of the ones above? The token costs are a big concern and one that a lot of people are talking . Tokenomics as they called it. I like the idea that the cost of a token is less relevant than the outcome that spending on that token created. If I spent $100 on a token that produces $500 in value, we should all want to spend on tokens. The challenge right now goes back to the point that many of these AI efforts aren’t producing the outcomes many hope to achieve.

This has become a really big topic in the world of healthcare AI. We need to give the agent access to be able to have the agent produce the outcomes we want. However, that access can lead to really poor outcomes, especially if we give that agent agency.

I like this other take on AI and speed. Humans actually make mistakes similar to AI because we’re not deterministic either. However, the speed at which we make mistakes is much slower. That’s a different challenge.

Are we overconfident? I think so.

As I often say, “Ignorance is bliss, until it’s not.”

My other reminder is that Shadow AI should be seen as a demand signal and not a reason for punishment. Satisfy the demand in an appropriate way rather than punishing the person who’s trying to do their job better and likely faster.

This could have gone in the security article, but I liked it in this AI article. If you don’t have a strong CISO, they’re going to be walked over by the people that want to implement AI. This could be a problem for many because the business outcomes can be compelling, but the security risks are real too.

Agent security is so new it does feel like a challenge. What do you think of the 4 steps listed in the image above as a place to start?

The reality we face right now is we need to focus our healthcare organizations on cost discipline and value economics. If you don’t have cost discipline, then your AI costs can go off the rails and stifle future projects. However, cost doesn’t matter nearly as much if the value that’s being created is more than the cost. Although in healthcare, that can be challenging. It can be tricky in healthcare to say that you want to generate a bunch of new revenue for your organization. There are some ways to do it, but it’s more challenging than in other businesses. That said, there should still be a value equation that every healthcare organization is considering as you implement AI. Healthcare doesn’t need another AI project, but it does have some outcomes it would like to improve and AI can help with those outcomes.

Thanks again to the SHI International team for having me out to their SHI Summit. It was an enlightening experience to hear from some great people and connect with real users from healthcare and other industries.

What did you think of these AI perspectives? Anything new that you hadn’t heard before? Any good reminders? Let us know on social media.