No. Your CRM, file shares and data rooms stay where they are, and they remain your systems of record. Metal reads from them and writes verified results back. It isn't a replacement store or a data vendor. Your teams keep working in the AI tools they already use, so there is no new destination to adopt and no firm-wide retraining.

A connector gives the model access to a system. It doesn't tell the model that Halden Corp in an email, HALDEN INDS. in your CRM and Halden Inc. in the CIM are the same company. It doesn't know which of three EBITDA figures your IC underwrote to, or why the firm made the call it did. So every question starts again from raw files, and the model rebuilds that structure on the fly. A context layer does that work before anyone asks. It resolves entities across your systems, ranks conflicting values by source, and keeps the firm's decisions on the record, current as your sources change.

Yes. Your data is yours. It is never used for training, not by Metal and not by any AI provider. Each firm runs in its own tenant, encrypted in transit and at rest and independently audited. Everything you add or build stays yours, and one request starts a full export, delivered within 30 days.

Metal ingests continuously as your sources change. When a revised model or an updated CIM lands, its figures join the record and are ranked against the existing values by source priority, confidence and recency. Overrides and values your team has pinned always take precedence. Earlier values are never deleted. They stay attached to the record with their own source, so you can see what changed and when.

Yes. Through Metal's native MCP server and open API, the skills, agents and workflows your team has built in Claude, ChatGPT, Copilot or in-house can read directly from your Context Graph. You can also build your own in Metal with Skill Builder and run them as Workflows, each one reading from the same resolved record.