Mili on The Customer Wins: Building AI agents advisors can actually trust
Chirag Gandhi joined Rich Walker on The Customer Wins to talk about what it takes to build AI that actually works inside a regulated industry. No bots in meetings. Notes that appear in real time. And a clear view on where human judgment always stays in the loop.
The conversation
Mili's Co-founder and CEO Chirag Gandhi joined Richard Walker on The Customer Wins podcast as part of the show's Emerging Tech Series. Over 30 minutes, they got into the design choices behind Mili, what agentic AI actually looks like in practice, and where the hard limits are.
Rich Walker built Quik!, a forms automation platform for financial services, because he lived the problem firsthand. That shared context made for one of the more honest conversations about AI in wealth management.
What was covered
Why Mili works silently in the background.
Mili never sends an AI bot to join a meeting as a third participant. The space between an advisor and their client is too important. A client might be navigating a divorce, a death, something they're only comfortable saying in a room with one person they trust. Mili listens quietly, with consent, and structures notes in real time while the meeting is still happening.
Notes that appear during the meeting.
Because Mili streams rather than records, the notes build as the conversation unfolds. Mili structures them, corrects itself, and keeps going. By the time the meeting ends, the notes are already there.
The three A's: Ask, Automate, Alert.
Agentic AI develops across three stages. Ask is where most tools live today retrieving information, answering questions across your tech stack. Automate kicks in with deep workflow understanding. Mili recently filled 1,400 forms for a single enterprise customer during a client migration. Alert is the hardest stage: doing something before you're asked. That's where Mili is headed.
Where human review always stays in the loop.
For anything client-facing, an advisor signs off. 80% of Mili's follow-up emails go out with very minor edits but a human still reviews and sends. Clients shouldn't get an email before they've walked to their car.
Build vs. buy, the honest version.
AI learns from data. The more context, the better the system gets. Mili has processed hundreds of thousands of advisor meetings. That depth compounds over time in ways that are hard to replicate internally.
A line that stuck
"The space between the advisor and the client is pristine."
That's the design philosophy behind everything Mili builds.
Listen to the full episode
🎙️ [Listen here → https://www.youtube.com/watch?v=Ur3L8Olv01Y&list=PLyqIzgRaYEkUK2La6qL_suqEF-imFLtCg&index=2