V7’s V7 Go platform turns corporate files and datasets into a structured, queryable Context Graph that AI agents can use directly. The system aims to prevent models from having to rediscover a company’s background on every call — for example, which fund report is current or how an entity is named across multiple systems.
Why this matters
Business context is often scattered across documents, data rooms, spreadsheets, emails and internal tools. That fragmentation causes repeated searches, missed information, and hallucinations — especially in finance, insurance and real estate, where retrieval accuracy is essential.
How the Context Graph works
- V7 Go connects to repositories such as SharePoint and Google Drive, scans files, and extracts entities, relationships, facts, attributes and metrics.
- It organizes that information into a graph that is faster and cheaper to traverse than long‑context approaches.
- Each fact is linked to a new or existing record and preserves cited evidence to the original source. If the graph lacks needed detail, V7 Go falls back to document search (RAG).
Models and role assignments
V7 combines multiple models for different tasks:
- GPT‑5.6 Luna handles high‑volume, structured extraction and the Context Graph ingestion.
- GPT‑5.6 Terra and GPT‑5.6 Sol power Workflows for chat, tool use and more complex multi‑step reasoning.
- V7 has begun using GPT‑6 Astra for the most demanding Context Graph queries.
V7 says that with context, models and tools working together, agents run 50–100 step workflows in minutes with 99.9% accuracy and an auditable trail for every decision.
Benchmarks and test results
V7 reports that the Context Graph reduces errors and hallucinations: on the HERB benchmark — which measures finding and connecting information across enterprise systems — V7’s retrieval‑only system outperformed the official baseline by 69% and reduced hallucinations on unanswerable queries by 38%.
To stress test graph querying, V7 built a tougher benchmark using messy, real‑world data across thousands of documents with four difficulty levels. On the very‑hard level, GPT‑5.6 Sol scored 78% while GPT‑6 Astra scored 89% accuracy. On the easy, medium and hard levels both models scored close to 100%.
Customer outcomes and efficiency gains
V7 lists several customer impacts from using the Context Graph:
- Asset managers can screen deals 21× faster, cutting a full‑day process down to about 15 minutes.
- A financial services team reduced review time from more than 100 hours to under 10, saving $12,000 in expert costs per task.
- Insurance teams lowered errors in claims processing by 13.5% compared with a manual baseline after granting agents access to the full history of prior claims and existing policies.
Cost and performance optimizations
- V7 measured a 78% lower cost per document using GPT‑5.6 Luna versus GPT‑5.4 mini.
- Moving document‑heavy V7 Go workloads from the Chat Completions API to the Responses API reduced token use by roughly 5% for some PDF‑heavy workflows and improved caching reliability.
- Key workflows that make several external calls now finish up to 50% faster in V7’s latest harness.
Workflow orchestration and auditability
V7 Go maps each workflow step to tiers (fast, medium, smart). The platform guides models through long‑horizon tasks, deterministic code steps, handoffs to smaller models, file generation and integrations while keeping an auditable trace of every run.
Integrations and developer tools
Context Graph ingestion and querying are exposed via V7’s MCP server, so customers can access it from ChatGPT and other compatible clients. V7 Go workflows can also be created through MCP in Codex. These simplifications reduced the time to create a medium‑length workflow from about one hour to roughly 20 minutes.
Next steps
V7’s longer‑term plan is to make shared memory proactive: workflows that trigger when facts in the Context Graph change, flag inconsistencies and identify analyses that need reinspection. For example, a restated fund report could prompt V7 Go to flag downstream work still using the old figures.
Closing comment
“Our goal is to help enterprises re‑tool for the age of AI, with workflows that solve mission critical tasks, and memory that outperforms us humans,” said Rizzoli. V7 argues that the finance firms that extract real value from AI will be those with the best context, not simply the most agents.
