V7

V7 Go deploys pre-built, domain-specific AI agents (Financial Agent, Legal Agent, Insurance Agent) that execute end-to-end workflows by chaining multiple LLM calls, document extraction, and analysis steps. Agents are instantiated within V7's infrastructure with configurable triggers (event-based activation via Zapier, API calls, or scheduled execution) and output routing to CRM systems, OneDrive, or data rooms. The platform abstracts multi-step reasoning chains behind a workflow UI rather than exposing raw API endpoints, enabling non-technical users to execute complex document analysis pipelines without prompt engineering.

V7 Go integrates with external data sources (PitchBook, Dealroom, data rooms, OneDrive) and event systems (Zapier) to automatically detect new documents and trigger agent workflows. Documents are ingested via API connectors or file upload, with metadata extraction (source, timestamp, document type) used to route to appropriate agents. Trigger logic supports event-based (file arrival), scheduled (daily/weekly), and manual (user-initiated) activation modes, enabling hands-off automation of document processing pipelines.

V7 Go provides real-time monitoring of workflow executions with status tracking (pending, running, completed, failed), execution duration metrics, and error logging. Failed executions are logged with error details and can be retried manually or automatically. Status updates are pushed to users via email notifications or webhook callbacks. Execution history is retained for audit purposes and performance analysis.

Enforces per-account token usage limits and quota management to prevent unexpected cost overruns. The platform tracks token consumption in real-time, alerts users when approaching limits, and stops processing when limits are exceeded. Administrators can set usage limits per account, team, or project; limits are enforced at the agent execution level. The system provides usage dashboards and reports showing token consumption by agent, document type, and time period.

Enables agents to execute Python code snippets for custom data transformations, calculations, or logic within extraction and processing workflows. Code execution is sandboxed and scoped; users can define Python functions that operate on extracted data and return results. The system manages code execution, error handling, and timeout enforcement. Available libraries are limited to a curated set (NumPy, Pandas, etc.); external API calls and file system access are restricted.

Automatically assesses document quality and processing readiness before extraction, identifying issues like poor image quality, missing pages, or unsupported formats that may impact extraction accuracy. The system provides quality scores and recommendations for document preprocessing (rotation, enhancement, OCR). Quality assessment is performed before agent execution, enabling users to filter or preprocess documents before processing.

V7 Go routes agent analysis results to multiple destination systems (CRM, OneDrive, data rooms) with automatic format transformation. Extracted data is mapped to CRM fields (deal records, contact enrichment), documents are stored in OneDrive with metadata tags, and summaries are pushed to data rooms for stakeholder review. Routing rules are configured per workflow, enabling a single agent execution to populate multiple downstream systems without manual export/import steps.

V7 Go provides token-level usage reporting and cost calculation, tracking LLM tokens consumed per workflow execution, document processed, and agent invocation. Token Reports dashboard displays usage trends, per-user consumption, and cost breakdowns. Pricing is volume-based (pay-per-document or pay-per-token processed) with custom pricing tiers per customer. Usage limits can be configured per user or organization to enforce cost controls and prevent runaway spending.

V7 Go implements role-based access control (RBAC) with customizable roles and per-user license management. Roles define permissions for workflow creation, agent execution, data access, and reporting. License tiers control feature access (basic agents vs. premium agents, read-only vs. edit permissions). User provisioning is managed through the platform's Account Management interface, with support for team hierarchies and department-level access controls.

The Financial Agent extracts and analyzes financial documents (term sheets, cap tables, financial statements, pitch decks) using domain-specific entity recognition trained on financial terminology and deal structures. The agent identifies key metrics (valuation, funding amount, investor names, terms), extracts cap table data, and flags financial red flags (mismatched valuations, dilution anomalies). Extraction is performed via multi-step LLM reasoning chains that parse both structured tables and unstructured text, with confidence scoring for each extracted field.

The Legal Agent analyzes contracts and legal documents by extracting key clauses (liability, indemnification, termination, IP ownership), identifying non-standard terms, and flagging legal risks. The agent uses clause-level entity recognition to locate specific contract sections, extracts clause text and associated obligations, and compares against standard templates to identify deviations. Risk flags include missing standard protections, unfavorable terms, and potential conflicts with other documents in the deal.

The Insurance Agent analyzes underwriting documents (applications, medical records, property assessments, loss history) to extract risk factors and generate underwriting recommendations. The agent identifies key risk indicators (health conditions, property hazards, claims history), calculates risk scores based on extracted factors, and flags cases requiring additional underwriting review. Analysis combines structured data extraction (from forms) with unstructured text analysis (medical notes, assessments) to provide comprehensive risk assessment.

V7 Go supports multiple trigger modes for workflow execution: event-based triggers (file arrival in data room, new PitchBook deal, Zapier webhook), scheduled triggers (daily/weekly batch processing), and manual triggers (user-initiated execution). Trigger configuration is defined per workflow, with support for conditional logic (e.g., only trigger if document type matches pattern). Execution is queued and processed asynchronously, with status tracking and completion notifications.

V7 Go automatically extracts and enriches document metadata during ingestion, including source system (PitchBook, Dealroom, OneDrive), upload timestamp, document type classification, and associated deal/company information. Metadata is used for routing (which agent to invoke), filtering (which documents to process), and tracking (audit trail of document origin). Enrichment includes linking documents to CRM records and adding contextual information from source systems (company name, deal stage, investor names).