AI First Steps: Four Lessons from Our Partners
Artificial intelligence is now part of how every organization delivers information, guidance, and connectionāor plans to. The question is not whether to use it, but when to start, how much time to invest, and how to embrace these tools responsibly rather than recklessly.
This post summarizes four essential learnings we developed as we've worked with our partners. None are purely technical; the challenge is always and also organizational.
Lesson 1. Guardrails Come First
One partner told us:
"Newly diagnosed patients are our No. 2 category of visitor, and they are likely new members. They aren't looking for detailed information ā they're looking for reassurance."
So we added engagement guardrails on top of our existing safety guardrails to help ensure that human connection happenes when it matters most.
A friendly chatbot is still often easier for some people to start with, and can find details people may miss. Yet it still serves as an on-ramp to peer groups and personal support.
For medically focused nonprofits, the lesson is to recognize what information is most valuableābut also when it's time to talk to a human.
Lesson 2. Build Cultural Readiness Early
Olivia, an RN at one of our partner organizations, was initially wary of an AI system that drafted replies she had always written herself. "I want to offload the boring stuff, but I'm not comfortable with it making clinical judgments," she told us.
Together we considered the landscape of medical, non-medical, and quasi-medical questions. We decided that anything both medical and even moderately serious should trigger a human referral, while quasi-medical questions could proceed with disclaimers and context. Within a week, Olivia said she could "focus better on the people who actually need my time."
Organizations that wait often find themselves playing catch-up. New technology will still be available later, but may feel alienāor have advanced so much that it becomes overwhelming. It's best to jump in now: people don't become comfortable by reading articles; they become comfortable by using the tools, asking questions, and adjusting the limits.
Concrete experiences build confidence and discernment from executives on down. AI literacy or conceptual comfort turns out to be as important as technical know-how.
Lesson 3. Engagement: For Better or Worse
A chatbot or virtual advisor deepens engagement when it helps people do something concreteāunderstand a bill, find a resource, or answer a question. It can also short-circuit engagement when it becomes a one-and-done quick-answer machine. Members get what they came for and leave, or head to ChatGPT and never visit the experts.
As we wrote in our earlier post on Engagement Drain: "Public AI tools are now remarkableābut there are human elements to every medical situation too."
Our turning point came when we realized AI's best answer is often to include a referral to a trusted resource, including our insurance experts or our partners' support or social-work staff. When AI "stays in its lane" by collecting and transparently organizing expert information, it drives more interaction on the front end and deeper trust on the back end.
Lesson 4. Learn in Public, and Soon
Introducing AI is not a technical milestone or (only) a website redesign component. It's an organizational capability that changes how people use new tools, think about safety and privacy, and empower members.
Our own team is learning from our partnerships as well. "I see it now. The best answers may still be when they call in. Our job is to answer the easy ones and make sure people show up already better informed."
Users can still return to our advice serviceāit's a continuing human-in-the-loop interaction.
AI is adding value in new ways so quickly that even we don't know where the journey ends. What we do know is that early, deliberate, shared learning is essential to prepare for the next steps. And those steps will be urgentāthis is not a trend to ignore.