Prototype
Built by me
Ekam
One app for the agent who sells both insurance and mutual funds, and juggles three apps to do it. Researched, designed and prototyped end to end, then paused on purpose.
FOR
Ground-level agents who sell both insurance and mutual funds
MADE WITH
Claude, Figma, Expo
TIMELINE
April 2026, about three weeks
ROLE
Solo: research, product and design, with Claude executing to my direction
Why I built it
It started with my father, who has spent about 30 years in the insurance industry, in senior and leadership roles with big companies and institutions, mostly on the sales side. Talking through how agents really work, one thing kept coming up: many hold both insurance and mutual fund licences, juggle two or three apps and reconcile commissions by hand. Companies were moving from one side into the other, but nobody was built for the agent who does both.
Who it’s for
The dual-licence agent
Paid on commission and busy with clients, switching between an app per insurer, another for funds, and a notebook for what they are owed.
Build log
From conversations with my father to a working prototype
Mar 2026
Conversations with my father
How agents earn, which licences they hold, and where their day leaks time.
Apr 2026
Research and benchmarking
Two research rounds and a benchmark of agent apps on both sides of the market.
Apr 2026
Idea validation
Scored 3.8 to 3.9 out of 5: worth pursuing, with refinements.
Apr 2026
PRD and design strategy
13 modules and 97 planned screens, plus a regulatory blueprint.
Apr 2026
Prototype, four versions
Onboarding, a command centre, full SIP and insurance journeys, CRM, commissions and a client app, on mock data.
Paused
Stopped before launch, on purpose
A regulated product needs licences, legal cover and a technical co-founder. Not something to launch alone.
After
The steps I skipped became skills
Referencing, benchmarking, idea validation: one skill for each step I’d rushed past, so I wouldn’t skip it again.
Decisions
A few calls I had to make along the way
Decision 01
The day starts with money to earn
An agent is paid on commission, so home opens on what they can earn today, not on a dashboard. Every signal is a real reason to call a client, with what it’s worth, and its detail says why it matters and what to do next.
Decision 02
One client, both sides of the business
The same client buys funds and cover from the same agent, but the tools split them into two apps. Here every client sits in one list, tagged with what’s missing, and the profile shows funds and policies side by side. A zero under Policies isn’t empty data, it’s the next conversation.
Decision 03
Commission you can plan around
Agents keep a notebook of what they’re owed because statements are hard to trust. So earnings are split by type with what’s still pending, and a 12-month forecast shows what’s driving the number: active SIPs, renewals due, new clients. They can change the assumptions and see it move.
Decision 04
The app starts from the licence you hold
What an agent can sell depends on their licence: mutual funds, POSP, IRDAI, or both. So onboarding asks that first, then fetches their registrations from IRDAI with their PAN. They confirm their agent codes instead of typing them in, and only see what they’re allowed to sell.
Behind the build
Compliance is part of the job, so it lives in the app: CPE hours, deadlines, and WhatsApp templates ready to send
Two desks in one app: funds on one tab, insurance on the next
Me and AI
It starts in my head, not in a prompt
Ekam was the first product I built with Claude end to end, and honestly, I was excited. I had the problem clear in my head, but I rushed past steps to get to a prototype, and that cost me round after round of iterations just to make the design consistent. This is where my way of working started to take shape.
Research and validation
01
S
Me
Came in with what my father had told me: how agents earn, the licences they hold and where their day leaks
AI
Claude
Brainstormed with me, then ran research, a benchmark and an idea validation
S
Me
Checked it against how agents actually work and decided it was worth building
Final callMe
PRD and strategy
02
S
Me
Set the scope around the dual-licence agent and what they earn
AI
Claude
Drafted the PRD, the module map and a regulatory blueprint
S
Me
Reshaped it, but I was moving fast and skipped steps to get to screens sooner
Final callMe
Prototype
03
S
Me
Jumped into the prototype early, excited to see it working
AI
Claude
Built it in Expo across four versions, on mock data
S
Me
Kept catching inconsistencies and iterating, which is where I learned that skipping steps doesn’t save time
Final callMe
Where it stands
I paused it at the prototype, on purpose
I had no one close enough to the industry to pressure-test it, and a product under SEBI and IRDAI needs licences, legal cover and a CTO-level partner. Launching alone would have made me personally liable. I’d pick it up again with the right partner.
What I learned
01
Getting things done with AI isn’t the problem. How you get them done is. I skipped steps to rush to a prototype and paid for it in iterations to get the design consistent.
02
A prototype answers whether something can be built, not whether it should be. I needed people from the industry earlier.
03
In a regulated space, the product plan has to start from the rulebook, not the screens.
Want the
full story?
I'm happy to walk you through the files, the trade-offs and what I'd change today. Or tell me about something you're building.
Message me ↗
Based in
Mumbai, working with teams anywhere