From data room to IC memo.

Underlying provides an end-to-end pipeline for high-stakes deal intelligence, combining the scale of AI with the accountability of a human review gate.

Data Room Ingestion

Connect directly to virtual data rooms or upload document sets. Underlying automatically structures, indexes, and normalizes financial, legal, and operational files for instant analysis.

Deep Extraction

Go beyond simple keyword matching. Our AI agents can synthesize context across hundreds of pages, extracting complex clauses, edge cases, and hidden liabilities.

Mandatory Review Gate

The core of our platform. Every AI-extracted fact must be explicitly approved or flagged by a human reviewer before it can be added to the final output.

Source Citations

Every finding is hyperlinked directly to the source document, page, and exact paragraph. No more hunting for where a number came from.

Audit Trail Logging

We log who ran the analysis, what the AI found, who reviewed it, and when it was approved. A complete, immutable chain of custody for your diligence.

Deliverable Generation

Once facts are reviewed and approved, instantly compile them into your firm's standard IC memo, red flag report, or diligence matrix format.

Free-form Chat & Investigation

Beyond structured review gates, Underlying offers a powerful Chat interface that allows your analysts to ask complex, multi-document questions using natural language.

  • Multi-model routing: Choose between OpenAI, Claude, and more depending on your specific task constraints.
  • Run entire workflows: Trigger complex diligence tasks directly from the chat prompt.
Are there any change-of-control provisions in the master service agreements that trigger upon a private equity buyout?
U
Yes. I analyzed 45 MSAs. 3 contain strict change-of-control clauses triggered by a PE buyout. See sources: [MSA_Vendor_A.pdf, Pg 12], [MSA_Vendor_C.pdf, Pg 4].

The Orchestration Engine

Underlying isn't a wrapper around a single LLM. It's a stateful orchestration layer that chains multiple models, API calls, and python code execution environments to solve long-horizon financial tasks.

Multi-Model Routing

Automatically routes sub-tasks to the optimal model—using OpenAI o1 for complex reasoning, and Claude 3.5 Sonnet for massive document ingestion.

Ephemeral Processing

Data is processed strictly in memory and wiped immediately after the workflow completes. Bring Your Own Key (BYOK) supported for ultimate control.

Confidence: 98%
"Change of control provision triggered upon 50% voting power transfer."
Auto-routed to final output
Confidence: 62%Human Review Required
"Liability cap is $5M, excluding gross negligence."

Confidence-Based Routing

AI shouldn't replace your analysts; it should augment them. Underlying uses confidence-based routing to automatically handle clear-cut extractions, while escalating ambiguous clauses directly to a human reviewer.

  • No Black Boxes: Ambiguous outputs are halted until a human verifies the source citation.
  • Self-Calibrating: Set custom confidence thresholds based on the risk profile of the specific deal.

Firm-Native Template Integration

Don't settle for generic text generation. Underlying's Forward Deployed Engineers map the AI outputs exactly to your firm's proprietary Excel models and PowerPoint templates.

Connectors: Dealcloud, Grata, Tracxn, Crunchbase, SEC/EDGAR, Webcrawl, Pitchbook, Factset, CapIQ, AWS S3, Salesforce, Slack, Sharepoint.

Excel
Financial Models & Red Flags
PowerPoint
Pitch Books & Board Decks
Word
IC Memos & Legal Summaries
DealCloud
Live CRM Pipeline Updates

Ready to upgrade your firm's diligence?

Stop relying on black-box AI tools that you can't trust. Bring accountability back to your workflows.