By Amber Bradley, Editor-in-Chief, TalkLPnews
When was the last time you were a beginner at something? Not “a little rusty,” but actually bad. Clicking the wrong buttons, asking questions you’re embarrassed to ask out loud, feeling your face get warm because someone half your age already knew the answer.
For most of us in this business, it’s been a minute. You spend twenty years building expertise and you get comfortable being the person everyone else calls. That’s a good place to be professionally. It’s a rough place to be when something new lands on your desk and asks you to start over at zero.
Sean Jackson from Panda Restaurant Group said something at APEX that I keep coming back to: “I’m surprised, still today, that there are people who just haven’t touched it.” He was talking about AI. His team has been running fraud scoring and predictive cost modeling across 2,500-plus locations for two years, and they went from about 20% to 80%-plus accuracy predicting claim costs without hiring an actuary.
Sean is not a data scientist. He opened a browser and started asking questions.
I don’t think skepticism is what’s slowing our industry down here. Almost everyone I talk to accepts that this is going to change how the work gets done. What’s actually slowing people down is not wanting to be the director who doesn’t know how something works. So we wait for the vendor demo, the corporate rollout, the training program somebody else is supposed to build. Meanwhile the analyst three levels down is already using it to triage exception reports on a Tuesday afternoon, and nobody told her she needed permission.
What it costs to find out
Ten minutes and zero dollars. The free tiers of ChatGPT and Claude handle most LP use cases, and if your company runs Microsoft 365, Copilot is probably already licensed and sitting there ignored.
Two rules before you start. Scrub personally identifiable information out of anything you upload to a public tool, because anonymized data gives you the same analysis with none of the exposure. Then check the output, every time. Sean describes AI as having a grad student on staff: fast, useful, occasionally confidently wrong. You’d never sign off on a grad student’s report without reading it first.
Somewhere to start
We put together three free downloads so you don’t have to build the on-ramp yourself.
- AI in LP: The Beginner’s Playbook. What AI actually does, which platforms to use, how to structure a prompt that returns something usable, and honest answers to the questions people ask me privately (starting with whether their data is safe).
- The LP Pro’s AI Starter Kit. Five real use cases pulled from Sean’s team, each mapped to a copy-paste prompt. Claims fraud scoring, cost modeling, report drafting, exception triage, and timecard analysis.
- Your First 5 AI Prompts for LP. One page. Print it, keep it next to your monitor, and stop staring at a blank prompt box.
Grab whichever one matches where you are right now. They’re free, they take about fifteen minutes to read, and there’s no vendor waiting on the other side of the download to call you.
My honest read on where this goes: the LP professionals who get fluent with this over the next 18 months will be the ones running departments, because they’ll be able to answer questions their peers can’t. Everyone else will still be waiting on a demo.
Pull one exception report this week. Scrub it. Ask AI what it sees. Give yourself permission to be bad at it for an afternoon.Download the AI guides at talklpnews.com, and join us at APEX in Nashville, September 27-30, where these conversa