Your Client Just Forwarded ChatGPT's Answer. Now What?
Will AI replace us — and how do we become AI-proof, and grow with it?
By Senthilkumar S Founder, AssetPower Finserve. 22+ years in general and health insurance.
It is 9.40 at night. Your phone lights up. A client has forwarded a long ChatGPT reply: a premium comparison, a waiting-period summary, even a neat table of “best plans.” Then comes the message: “Do I still need you for this?”
If that tightness in the chest is familiar, you are not alone. AI will take parts of our work. It will not take the work that earns trust. The advisors who grow will let the machine do the drafting, and keep the human part for themselves.
The strange truth about what machines find hard
In 1988, the roboticist Hans Moravec noticed something that still helps us. Computers could already do well on intelligence tests and play checkers. They still struggled with what a one-year-old does easily: see, reach and move. Calculation and a first draft are often easy for a machine. Reading a room, sitting with a worried family, deciding what not to sell — that is much harder to automate.
That is why a chatbot can produce a quote in seconds, but it cannot stand with a family at a hospital billing counter at midnight.
What AI will take — and what stays human
If we pretend nothing will change, we will be overtaken. If we assume everything will go, we will undersell our own value.
What AI will take first: first-cut quotes and comparison tables; standard answers on waiting periods, exclusions, SIPs and top-ups; drafts of emails, WhatsApp messages and renewal reminders; paperwork checklists and first summaries of policy wordings.
What stays human: trust, when a family is about to sign a health policy or start a long-term SIP; claims support at the hospital — explaining a cashless query, calming a spouse, calling the insurer, staying until a query is closed; behaviour coaching when markets fall, or when a neighbour shouts about an “AI stock”; family situations no template can hold — a wedding, a parent’s surgery, a difficult nominee; the courage to say, “This product is not for you”; and accountability when a claim is delayed. A chatbot does not answer that phone.
Palantir CEO Alex Karp said at Davos in January 2026 that AI “will destroy humanities jobs,” and that a general degree, without another skill, “is going to be hard to market.” He was more hopeful about people with practical skill, who would become “very valuable, if not irreplaceable.” We are not electricians, but the parallel is real. The playbook part of our job — a script, a standard comparison, a copied brochure — is what a machine learns first. The part that cannot be reduced to a template is the part that will pay.
Put your own job through Swadia’s TLC test
Sandeep Swadia offers a simple filter. He calls it TLC: Trust, Learn and Contact.
Trust. Does the work carry accountability that people will not yet hand to a machine? For us, yes. A family will not give a chatbot the last word on their health cover, their child’s SIP, or a stuck claim.
Learn. Can AI learn the task quickly? Yes, if the work already lives on a screen, follows a pattern, and a wrong answer is easy to spot. Quotes, FAQs, renewal lists and first drafts sit in this box.
Contact. Does the work need you to be there in person, not just behind a screen? Not every day. But the days that matter do: the hospital corridor, the kitchen-table review, the visit after a death in the family. That contact is the job.
Score your week against these three. The more a task is only “Learn,” the more you should let AI draft it. The more it needs Trust and Contact, the more you should show up in person. That is how we become AI-proof: not by ignoring the tool, but by refusing to become one.
A practical AI checklist — and one hard caution
Use AI as an assistant, not as a substitute for your advice.
Draft a WhatsApp or email, then edit it in your own voice before you send
Prepare renewal reminders and a short list of pending documents
Sketch an explainer post on waiting periods, nominees or staying with an SIP
Ask for a plain-language summary of a policy wording — then read the wording yourself and correct the machine
Build a first-cut comparison table, then check every figure against the insurer or the scheme document
The caution is not optional. Do not paste a client’s name, medical details, policy number, PAN, Aadhaar, bank account or claim papers into a public AI tool. Treat that as you would treat leaving a file in a tea shop. Use only anonymised examples.
Seven things to do this month
List your last 20 work tasks. Mark each one T, L or C. Hand the pure “L” tasks to AI for a first draft.
Choose one AI tool and practise it on a real client task every day, or a few times a week. The first attempts will be clumsy; that is how you learn.
Catch up on AI developments once a week, not every hour. And do not forward every alarming headline to clients.
Write down how you will show up when it matters — a hospital claim, a sharp fall in a SIP — and keep that promise.
Verify every policy figure, exclusion and fund fact the machine gives you. You own the advice, not the chatbot.
Keep client data out of public AI tools. Full stop.
Treat the next “Do I still need you?” message as a conversation, not a defence. Show what you will do that the screen cannot.
As Sandeep Swadia puts it: “You cannot control the ocean. You can only steer your boat.”
AI can draft the quote. It cannot sit with a family at the hospital, or hold a hand when the markets fall. The advisors who will be replaced are the ones who only ever did what a chatbot now does for free. The ones who will grow use the machine — and still show up.
Senthilkumar S 99625 47079
Ideas drawn from Hans Moravec, Alex Karp and Sandeep Swadia.
Comments