She asked her AI first. Then she asked her advisor.
Lisa Ellison spent most of an evening talking with ChatGPT about retirement. The next morning she forwarded the whole conversation to her advisor with one line on top: “ChatGPT says we should both wait until 70 to claim Social Security. Is that right?” It landed in the Ellisons’ household in Helmsted, next to everything their advisor already knew about them.
1. “Lisa sent this from her AI. What didn’t it know about them?” Helmsted read the full conversation against the household. The chatbot had treated them as two similar earners. They aren’t: Lisa’s own benefit at 67 is $1,100, and her spousal benefit brings her to $1,700, an amount that stops growing at 67. Dan’s larger check is the one that continues for whichever of them lives longer.
2. “Check our old CRM for anything on their health, family history, or when Dan plans to stop working.” Helmsted asked to use the CRM’s site, and a browser window popped up on its sign-in page. The advisor logged in himself, and Helmsted took it from there: it searched the Ellisons’ 62 notes going back to 2014 and brought back the three that matter. Dan plans to stop working at 68. His parents lived to 93 and 96, and Lisa’s mother is 88. And they’d rather not draw on the IRA before Dan stops working. It added all three to the household.
3. “Make an Interactive visual showing whether waiting until 70 is right for them, and what I’d recommend instead.” For Dan, waiting takes his check from $3,400 to $4,216 a month, $816 more for life, and with long lives on both sides, that larger check for the survivor is worth even more. For Lisa, three more years of waiting gives up $61,200 and adds nothing. The recommendation: Dan at 70, Lisa at 67, both checks starting in the same month, four years from now. Two sliders let them set how long each of them lives and see when waiting pays.
4. “Draft my reply to Lisa.” It opens by thanking her for sending it, answers her question in one sentence, and links the report in the client portal. It went out before lunch.
Their AI conversation is a gift
Clients are already talking with AI about their money, often at length and late at night. Those conversations are a record of what they worry about, what they’ve read and what they want. The chatbot’s answer is usually sound in general. What it lacks is the household. Bring the whole conversation into Helmsted, let it find what the chatbot didn’t know, and answer the client about their own plan. The technology does the comparison. The advisor makes the call. More on this in Wednesday’s article, When a client brings an AI recommendation, make it personal.
Getting creative with materials with Helmsted
Start with their words. The report opens with what Lisa sent, so the answer reads as a reply to her, not a correction of her chatbot. Two people, two answers. The same advice is right for Dan and wrong for Lisa, and side-by-side bars make that clear at a glance. Let them test it. Delay is insurance for a long life, so the sliders put the lifespan question in their hands. Say what would change it. A change in health, Dan stopping work early, or needing income the portfolio shouldn’t carry. Naming those now sets up the next review.
One thing makes all of it work: context. Helmsted needs both spouses’ ages, each benefit estimate at full retirement age from their Social Security statements, spending, the accounts that will carry them until they file, and anything they’ve said about health and family. If that history lives in an old CRM, Helmsted can now go read it for you.
New this week: browser control, tell Helmsted to go to any website to complete a task.
