Project Overview

Our client needed compliant, explainable RFP/DDQ/QBR responses that work with Loopio but go further: they needed to pull from CRM and policy docs, enforce governance (HITL/SoD), create curated Best-Answers datasets, and automate downstream handoffs (CRM/email) inside the client’s VPC.

Solution: Odyssey 3.0 Agent Fleet

A private, governed Agent Fleet on Odyssey 3.0, orchestrating MCP (Model Context Protocol) tools for retrieval, drafting, disclosure checks, and governed delivery (CRM/email) with audit trails and SourceSnips citations.

Timeline

  • Next-day prototype deployment
  • POC in 1–4 weeks.

Challenges & Objectives

  • Private AI & Data Isolation (IP Risk)
    • Objective: Ensure no training on your data; no cross-customer fine-tuning.
    • Solution: Client-dedicated instances in your VPC/on-prem; prompts, embeddings, and logs remain inside your boundary. (Odyssey 3.0 + Agent Fleet architecture.)
  • Governed End-to-End Workflows (Reliability/Bottlenecks)
    • Objective: Go beyond Loopio’s project strengths to enforce policy at runtime.
    • Solution: Control Tower enforces HITL approvals, SoD, and tamper-evident logs; MCP connectors orchestrate governed CRM write-back and approved email delivery.
  • Curated Content Library
    • Objective: Make responses consistent and explainable.
    • Solution: Curate and version a firm-specific Best-Answers dataset (by product, mandate, region, regulator) and bind workflows with Dataset Targeting for consistent, explainable reuse.

Private KB (one-liner): Odyssey maintains a company-owned private knowledge base behind the agent, scoped to approved datasets.

Odyssey 3.0 Architecture & Process

Where It Fits Next to Loopio

Loopio remains your RFP project system and Answer Library UI. Odyssey becomes the private KB + governed agent engine that:

  • Builds/maintains the curated Best-Answers dataset behind the scenes.
  • Pulls from Loopio/CRM/storage, adds citations and governance, and can write back to CRM or generate governed emails.
  • Works even when no marketplace app exists—via API (where permitted) or by staging approved exports in Snowflake/S3.
How It Works (High Level)

BUILD

  • Connect datasets: Loopio Library exports + CRM objects + policy/marketing docs.
  • Design workflow: Visual Agent Builder steps RAG, Compile, Structure, MCP Discovery, MCP Direct, Custom Tools.

GOVERN & DELIVER

  • Set thresholds (HITL/SoD) & trace everything (citations/history).
  • Publish downstream: push content back to CRM, generate client-ready PDFs/Docs, or send governed emails.
“Every outbound answer is explainable with SourceSnips and dataset provenance, ensuring both compliance and clarity in client communication.”

Tangible Benefits:

✓ Faster first drafts: Minutes instead of days, with citations and disclosure checks.
✓ Fewer compliance loops: HITL gates + SourceSnips provenance embed compliance upfront.
✓ Better reuse: Curated Best-Answers scoped by product/mandate/region.
✓ Operational fit: Keep Loopio as the UI; add governed agents for CRM/email handoffs.

InteliGems Value-Add Comparison:

 CapabilityLoopio (baseline)InteliGems Add-on
Answer Generation AI drafts from Library; private LLM endpoint; “no training on your data” stated.Drafts from curated, versioned Best-Answers bound via Dataset Targeting; multi-agent reasoning; full explainability.
CRM/EmailCRM reporting/status integrations; HubSpot example includes one-way sync; email notifications only. Governed write-back to CRM and approved email workflows via MCP (HITL/SoD, templates, audit trails).
Data Access Public integrations & Unleash connector (80+ sources). API ingestion where permitted, or warehouse/object storage staging (e.g., Snowflake → S3) under your control.
PrivacyPublished privacy/DPA; GenAI under “strict data protection” protocols. Private deployment (VPC/on-prem); no cross-tenant learning; full log/embedding isolation.
GovernanceProject/task governance; platform notifications. Control Tower: runtime policy enforcement, SoD, approvals, evidence export.

Get started today

Ready to Implement Governed RFP Intelligence?

Next-day prototype on a live RFP, then POC in 1–4 weeks with acceptance tests.