Many healthcare organizations have jumped on the AI train over the past several years. However, in many cases, the train often jumps the track and stalls out on the journey. Why? 

Four healthcare executives at the recent Becker’s Annual IT + Revenue Cycle Conference discussed core concepts that can lead to a successful AI selection process and implementation. The executives included:

• Drew Smith, chief data and analytics officer, ChristianaCare
• Kalyani Gopalan, executive director of analytics at Presbyterian Healthcare Services, Albuquerque, NM
• Alvin Liu, MD, director of an endowed AI center and leader of AI implementation and governance at Johns Hopkins School of Medicine
• Darrell Keeling, chief technology officer at Bronson Healthcare

The 5 concepts include:

• Defining the type of AI desired (generative vs. classical machine models) to determine the right type of ROI (clinical, operational, or financial). For example, for non-clinical AI uses, improvement in operational productivity is most often seen, but this can also improve financial KPIs.
• Monitoring hidden costs and having a sustainability plan. Costs that are often missing from the original AI proposal are:
  • Licensing costs per user
  • The cost of tokens (charged for every query sent to an AI model)
  • Additional technical labor needed after “go live” to review, optimize, and rebuild the code.
  • Failed pilot efforts
• Management of user adoption and culture. No matter how good the AI platform, if it’s not used, it’s a cost center, not an asset. Clinicians need to trust clinical applications. They won’t use it if they think it can cause harm. Pilot programs should be looked at as a learning exercise about user, not just the technology involved. If an AI model isn’t working for the organization, it’s time to admit failure and move on to something else.
• Considering the cost of inaction. Most organizations don’t want to move too fast with an AI implementation, but moving too slowly is also a risk.
• Addressing AI as an empowering tool, not a replacement. There is a fear that AI will replace jobs and therefore people. While AI does help to automate many tasks, it also frees up staff to upgrade skills and improve performance.