Live App →

Digital Advisor

Conversational AI interface that guides wealth managers and clients through portfolio insights, document uploads, and financial decision support.

Digital Advisor

Scope — The primary conversational interface for the Sentinel platform, enabling natural-language interactions across portfolio analysis, document processing, market research, and operational support with enterprise-grade guardrails.

Executive Summary

The Digital Advisor is the face of Sentinel’s AI capabilities. Built on the Zen Chatbot Platform and orchestrated by the Agentic Backend, it provides a unified chat experience where users can upload documents, ask portfolio questions, request reports, and receive guardrail-protected financial guidance.

Unlike generic chatbots, the Digital Advisor is deeply integrated with the platform’s backend services: every conversation can trigger document extraction (Nexus), portfolio calculations (Agentic), live market search (Web), and support ticket creation (Integrations).

The Problem

Traditional wealth management software is form-driven and static. Advisors must:

  • Navigate 10+ screens to build a client portfolio view
  • Manually copy data between document extracts and CRM notes
  • Wait hours for operations to generate custom reports
  • Rely on generic web search for market updates

Clients, when given portal access, find the interface intimidating and rarely self-serve beyond balance checks.

The Digital Advisor replaces this with a single conversational thread that understands context, remembers history, and takes action.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                      Digital Advisor                             │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────┐      ┌──────────────┐      ┌─────────────┐    │
│  │   Sentinel  │─────▶│   Studio     │─────▶│    Zen      │    │
│  │   Frontend  │      │  Middleware  │      │  Chatbot    │    │
│  │  (Chat UI)  │      │  (Auth/BFF)  │      │  Platform   │    │
│  └─────────────┘      └──────────────┘      └──────┬──────┘    │
│                                                     │           │
│                              ┌──────────────────────┘           │
│                              ▼                                  │
│                       ┌──────────────┐                         │
│                       │   Agentic    │                         │
│                       │  Supervisor  │                         │
│                       │  (Intent +   │                         │
│                       │   Routing)   │                         │
│                       └──────┬───────┘                         │
│                              │                                  │
│         ┌────────────────────┼────────────────────┐             │
│         ▼                    ▼                    ▼             │
│  ┌─────────────┐    ┌──────────────┐    ┌─────────────┐      │
│  │   Nexus     │    │   Portfolio  │    │   Web Search │      │
│  │  (Document) │    │   Review     │    │   + Guardrails│     │
│  └─────────────┘    └──────────────┘    └─────────────┘      │
└─────────────────────────────────────────────────────────────────┘

Conversation Flow

  1. User sends message (text, image, or document)
  2. Zen validates session and enriches with history
  3. Agentic Supervisor classifies intent (guardrail + supervisor run in parallel)
  4. Router dispatches to the appropriate agent:
    • document_qa → KB retrieval + RAG
    • portfolio_review → Portfolio analytics engine
    • general_chat → Web search + reasoning
    • vision → Vision model analysis
  5. Response streamed back via SSE with citations and action cards
  6. Session state updated for context retention

Personas & Journeys

Relationship Manager (Primary User)

  1. Starts a new chat: “Analyze Mr. Sharma’s portfolio”
  2. Uploads Sharma’s latest CAS statement
  3. Asks follow-up: “What’s the sector concentration?”
  4. Requests: “Generate a rebalancing report”
  5. Receives PDF report with AI commentary, charts, and PII-masked summary
  6. Shares report link with client via integrated Outlook

Client (B2C Self-Service)

  1. Logs into white-labeled advisor portal
  2. Asks: “How am I doing this quarter?”
  3. Digital Advisor retrieves latest portfolio data and computes Q-o-Q change
  4. Explains in plain language with visual chart
  5. Client asks: “Should I invest more in debt?”
  6. Advisor responds with educational content + risk profile reminder (no direct advice)

Operations Manager

  1. Receives alert: “5 extractions failed in last hour”
  2. Opens Digital Advisor admin channel
  3. Queries: “Show failed jobs for tenant Acme Wealth”
  4. Gets structured table with error codes and document IDs
  5. Drills into specific job and triggers reprocessing

Key Features

Feature Detail
Multi-Session Management Persistent chat history with auto-generated titles
Document Upload in Chat Drag-and-drop PDFs scoped to the current session
Streaming Responses Real-time token delivery via SSE for natural feel
Action Cards Embedded buttons for “Export Excel”, “Schedule Review”, “Create Ticket”
Guardrails SEBI disclaimer, no buy/sell recommendations, PII masking
Multi-Modal Text, image, document, and voice-ready architecture
White-Label Ready Themable UI for B2B/B2B2C partner deployments
Webhook Integration Nexus completion callbacks auto-update chat context

API Surface

Method Endpoint Purpose
POST /api/v1/chat/ Create session
POST /api/v1/chat/{id}/messages Send message
POST /api/v1/chat/invoke/stream SSE streaming
POST /api/v1/chat/{id}/documents Attach document
POST /api/v1/orchestrator/invoke Advanced agentic invocation

Security, Compliance & Operations

  • Guardrail Layer — Every response passes through a guardrail node checking for prohibited advice, PII leakage, and regulatory compliance
  • SEBI Disclaimer — Automatically appended to any response touching investment topics
  • Session Isolation — Chat context and uploaded documents never leak across sessions or tenants
  • Audit — Every message, attachment, and model invocation logged with full metadata
  • Rate Limiting — Per-user message throttling via Studio Middleware