Why I built ResolveIQ
The role asks three questions and says “show us you can do it.” Here are the answers - each a one-line claim, then the proof rendered as a diagram, a comparison, or a plan - with a working console behind them.
1 · What would you build on Lyzr for BFSI?
ResolveIQ - autonomous payment-dispute & chargeback resolution, as a new pillar in Agent Amadeo. Six specialized agents under a manager, sequenced by deterministic HybridFlow so the Reg E & network-deadline logic is code, not a hopeful prompt. It even refuses to act on a $1,310 friendly-fraud case and hands it to a human.
2 · Why is it the right use case?
Because disputes are where agents win the argument outright. A ~$15B, double-digit-growth problem with hard deadlines and evidence across five systems - irreducibly multi-step, deadline-bound, and human-in-the-loop. RPA breaks; a chatbot can't act. And it isn't an orphan bet: Lyzr published the dispute playbook + six-agent architecture, and Accenture invested to push agentic AI into banking.
| Can it… | RPA | Chatbot | Agent |
|---|---|---|---|
| Reason on ambiguous evidence | |||
| Act across 5 systems | |||
| Hold a Reg E / network deadline | |||
| Leave a replayable audit trail |
3 · How would you position it & take it to market?
Not as another agent builder - as the production-and-compliance layer for BFSI agents. Everyone can demo a dispute agent; ResolveIQ ships because explainability, tiered autonomy, and data sovereignty are native - built so the CRO and Model-Risk sign off instead of block. That's the exact gap where ~95% of GenAI pilots stall (MIT 2025) and 40%+ of agentic projects get cancelled by 2027 (Gartner). Five moves:
Co-sell into Accenture's BFSI accounts where the MSA & trust already exist. Neutralizes “will you exist in 3 years?”
Runs in the bank's VPC on masked data. Exit gate = measured result + named sponsor + budget line.
Platform + VPC usage ($0.03/run), then outcome pricing per dispute auto-resolved. Framed to the COO, never as “agent runs.”
VPC/on-prem, “we don't train on your data,” audit-replay spec + Responsible-AI one-pager. In BFSI, blockers stop more deals than buyers start.
By use case (disputes → KYC → AML → loan servicing) and by BU (retail → SME → wealth, or insurance claims). Same plumbing.
The asset is live - not a deck.
Run a real dispute through the six agents, watch it auto-resolve, and watch it stop for a human.
Open the console