Volume III · Fintech · Real-time communication
The Escalation Path
A real-time support system for a financial platform, built so that every user reaches the right person, and the product teaches itself.
- Problem
- Users had to reach the right broker fast, without repeating themselves.
- Outcome
- Live support chat with escalation, direct broker chat, and an in-app assistant that performs actions.
- Principal
- Scale App ↗
- My role
- Lead Flutter engineer
- Period
- 2025 — 2026
- Tech used
- Flutter · Dart · WebSockets · FastAPI · LLM agents
Chapter 1 · The problemThe Thesis
Support systems do not fail at the answer. They fail at the handoff. The real question is never whether a user receives a response, but whether they reach the right person before they give up.
Chapter 2 · The client & my roleThe Catalyst
Scale App connects users with brokers. The product needed communication that carried weight: secure, live, and accountable.
I built a real-time support chat with escalation workflows and secure broker connectivity, integrated account-linking with direct broker chat and monthly and yearly review flows, and developed a guided assistant that teaches the platform and carries out user-requested actions inside the app.
Chapter 3 · The hard partThe Friction
Chat is easy; routing is not. A conversation that starts as a simple question can become a matter for a licensed broker, and the system has to know when, and move the user without losing context.
The decision was to treat escalation as a first-class state of the conversation, not an afterthought. The same thread can begin with the guided assistant and end with the user's own broker, with nothing repeated and nothing lost.
“A support system is measured by its handoffs, not by its replies.”
Chapter 4 · The resultThe Monument
Users move from self-service to a human broker inside one continuous, secure channel, and recurring reviews are built into the relationship rather than left to memory.
- Live
- Support chat with escalation workflows
- 2
- Review cadences: monthly and yearly
- In-app
- Assistant that executes actions, not just answers