Origin
A prototype that helps private bankers estimate prospects' equity, run compliant research and draft a first outreach email, with a human reviewing every step.
Project Details
About the Project
Banks spend a huge amount of effort finding new ultra-high-net-worth (UHNW) clients. I built Origin to explore how AI could help with that, responsibly.
Rationale
At an SMU fireside chat, a guest speaker told me that one of the biggest problems facing banks is prospecting and originating new UHNW clients for private banking. I had never thought of AI in that role, and decided to build an app to help with it.
What Origin Does
Origin follows one clear flow for an advisor:
- bring your own data or try a sample — a prospect list and company details, or a small fictional sample;
- model equity — an estimate of each person's after-tax equity as a range with a confidence level, using vesting, tender offers and tax assumptions;
- research on request — a dossier with a compliance status, including sanctions and PEP screening; and
- draft the first email — a warm, low-pressure note for the advisor to review, which always ends with a transparency and opt-out line and is never sent automatically.
The public demo uses synthetic dossiers only, so it never researches real people; the AI writes the email draft.
Privacy First
I researched how different countries handle privacy and data security, and built those principles into the app. Data is read in the browser and never uploaded, uploaded files are never run as code, and a human reviews every step.
Origin is a prototype: its estimates are modelled, not verified net worth, and it supports a firm's compliance process rather than replacing it.
Project Satisfaction
My proudest moment was researching privacy and data-security practices and building them into the app. Being ethical and putting client data security first should never be compromised.
Biggest Challenge
Trying not to build everything. With AI it's tempting to keep adding features, and I lost weeks to that. A half-finished tool with ten ideas is worse than a working tool with one clear flow, so I focused on the smallest end-to-end journey: upload data, see an estimate, run the research, draft the email.
Key Takeaway
AI can play a role in places you don't expect. I wouldn't have seen this use case without the guest speaker's suggestion, and talking to people from different fields really widened my view of what AI can do.