Artificial intelligence (AI) was one of the big themes of the recent annual convention of LeadingAge, the association of non-profit retirement communities. There were at least 10 educational sessions about AI, plus a keynote speech.
Given that I am very interested in AI, I could have spent my entire time at the LeadingAge event immersed in all the AI content. But I felt I couldn’t afford to do that, so I selected a handful of the most promising-sounding sessions to attend. I wanted to know how AI was being applied in retirement communities.
I’m sure you remember the story of the blind men and the elephant. Each of them described the elephant in completely different terms, depending on whether he had felt the elephant’s skin, tusks, trunk, or tail. Well, that’s a good metaphor for the impression of AI use in retirement communities that I took away from the LeadingAge convention. The reports on AI use were all over the map, and it wasn’t possible to form a coherent overall picture. But there were some approaches and trends that I found interesting.
Setting the rules. Every organization that wants to adopt AI needs to set some ground rules, and retirement communities are no exception. In fact, the rules are even more important in retirement communities because they often deal with sensitive health information. The most basic requirement is to get a license to use a specific AI software package and insist that staff use only that package.
A license covers aspects of AI usage such as: how data that the AI software works with can be used, who owns the outputs of the AI software, what liability limits there are in case of problems, and so on. (Kendal-Crosslands has licensed Microsoft CoPilot in this way.)
Individuals using a product such as ChatGPT don’t need an explicit license, but organizations do. Without licensing, it is possible that the internal data the AI software is working with could make it out onto the internet, either directly or through its use as training data for future software versions. A license can prevent this and other potential problems—but staff must resist the temptation to use a different AI package that they like better based on previous personal experience.
Training. Several presenters spoke of the need for careful training of staff in the appropriate use of AI. It’s probably alright to use AI for a quick draft of an email—but you can’t blindly trust the result. You need to review it carefully before hitting Send. More complicated tasks require more supervision and, often, requests for revision. One CEO said of AI-produced output: “I never accept the first version.” Asking AI for a sophisticated program to automate what is apparently a relatively routine task can be asking for trouble and should be left to the IT department.
How ambitious should you be? One presenter made the distinction between “Individual & Workforce” and “Enterprise & Transformational” uses of AI. The former can be implemented by individual staffers who find ways to use AI in their normal daily work. They can then share their experience with colleagues. The targets of this approach might be routine documents of various kinds.
The “Enterprise & Transformational” projects are going to involve the IT department and perhaps outside consultants. Implementing this type of project is going to require data, governance, and change-management efforts. But such a project could have much broader impact on the organization and its mission.
One presenter suggested having staff members record a “dynamic journal” by talking to an AI software “agent” (using a cell phone) while going through their working day. The agent could summarize the day and might be able to suggest improvements. Another suggested that staff could ask themselves, “what part of my job gives me joy?” and then see whether AI could help with the rest.
Presenters were clear that staff would still need to be responsible for the quality of the data being collected. AI could use the data, but only if a human was verifying its accuracy.
Some AI-based products brought in from outside the retirement community are already proving valuable. One presenter mentioned a tool for moving through the estate-settlement process, a multi-step process that goes far beyond what is in the will. Another mentioned a tool that helps the marketing department prepare materials for eight different formats (print, web, and several social media formats) from a single batch of text and images.
Do you use a consultant? Consultants were involved in several projects described at LeadingAge. Implementing an ambitious AI project is often fraught with unanticipated difficulties that consultants may be able to address.
One retirement community had a consultant develop a movie-management bot. Movies were supposed to run every evening at 7 p.m., but that required a staff person to be free at that hour, consult the list of scheduled movies, and initiate the movie from whatever service offered it. The movies were often late or skipped entirely. The consultant located an independent streaming library where movie requests could be initiated by a computer program. A bot was developed that ran each evening, reading the week’s movie schedule, retrieving and playing the required movie, and updating a log file showing that the movie had run.
Another community had a consultant review a wide variety of potential uses for AI. Three were selected for initial implementation: an automated daily census report, a tool for plain-language answers to questions on community policies, and a tool for drafting communication to families of residents.
I came away from these sessions with an understanding that AI implementation is at a very early stage in retirement communities, and different communities are approaching it in different ways.
Although it was not mentioned much in the presentations, there was an undercurrent of uncertainty about where AI might lead. How trustworthy will AI turn out to be? Will there be layoffs because of AI? Will AI have a positive return on investment? Will some of the vendors go out of business? Is federal regulation coming, and if so, what impact might it have?
There’s a lot of buzz around AI; but it also feels like the Wild West, with the rules being made on the fly. We’re still trying to figure out what the whole elephant is like.
