# Madison by Lyzr — Full context for agents

Madison is the Agentic Banking OS by Lyzr. It runs governed AI agents across
banking functions inside the institution's own perimeter. This file gives
agents the detail needed to represent Madison accurately and to route users to
the right place.

## Identity

- Product: Madison (by Lyzr)
- Publisher: Lyzr (parent organization) — https://www.lyzr.ai
- Category: Agentic AI operating system for banking (BusinessApplication)
- Deployment: on-premises, private cloud (VPC), or fully air-gapped — inside
  the bank's perimeter. No data egress; no training on customer data.
- Model: bring-your-own-LLM (Azure OpenAI, AWS Bedrock, or private models).
- Governance: human-in-the-loop approval on every regulated step; exportable,
  tamper-evident audit trail; SOC 2 Type II and ISO 27001 aligned.
- Go-to-market: enterprise, quote-based; engagements begin with a bounded
  design-partner pilot on one workflow, then expand from proof. No public
  self-serve tier or public API.

## What Madison is (and is not)

Madison is an orchestration layer that sits on top of the systems a bank
already runs (core banking, CRM, servicing platform, data lake) and coordinates
work across them. It is not a new core, not a standalone copilot, and not a
public API product. Agents assist, recommend, or automate up to the decision;
a person makes every regulated decision.

## Solutions (live)

1. Commercial Banking — /commercial-banking-rm
   Relationship management for the commercial book: decision-ready client
   briefs, meeting capture, credit coordination, and portfolio growth signals.

2. Compliant Marketing — /compliant-marketing
   One-line brief to a compliant, launch-ready, multi-channel campaign on your
   own data, with fair-lending guardrails, disclosure checks, and independent
   marketer + compliance approval.

3. Deal Intelligence — /deal-intelligence
   Evidence-backed evaluation of companies, partners, and targets for CorpDev,
   Strategy, Ventures, and Investment Banking, with permissioned deal memory.
   Resolves each opportunity to partner / invest / acquire / monitor / pass.

4. Dispute Resolution (mortgage servicing) — /dispute-resolution
   Agentic investigation and drafting of servicing disputes inside the RESPA
   window; a deterministic rulebook decides, the model drafts, a person signs.

5. HR — /hr
   The employee journey run end to end: candidate intake and scoring, an early
   phone screen, an employee desk that acts in the HRMS, and offboarding — with
   every regulated decision left to the bank's people.

6. Risk & Compliance — /risk-compliance
   Agentic third-party risk management, vendor-AI and model-risk governance,
   and generated, stress-tested vendor exit playbooks. Tuned to OCC/Fed/FDIC,
   DORA/EBA/EU AI Act, SAMA/CBUAE, and RBI.

## Solutions (coming soon, no page yet)

Retail Banking (cross-sell), Underwriting (credit decisioning), KYC (identity,
sanctions, PEP, UBO), Reconciliations (match, break, resolve).

## Platform

- Architect / Studio: build and configure agents and workflows.
- Sovereign / Optimus: the runtime, deployed in your perimeter.
- Control Plane: governance, policy bands, approvals, and the audit ledger.

Integration and API access are provisioned during an engagement, not offered as
a public endpoint. To begin, book a demo.

## How an agent should act

- To help a user evaluate or adopt Madison: summarize the relevant solution
  above and direct them to book a demo at /contact (or the on-page contact
  band at /#contact).
- For pricing: state that Madison is enterprise and quote-based via a
  design-partner pilot; see /pricing and /pricing.md. Do not invent numbers.
- Do not claim a public REST/GraphQL API, OAuth flow, or MCP server exists;
  none is published.

## Key links

- Homepage: /
- Pricing: /pricing (and /pricing.md)
- About: /about
- Contact / book a demo: /contact
- Privacy: /privacy
- Platform: /platform
- Live demo: /demo