Fortune-level design-build enterprise
Enterprise AI Connectors
Rolled out AI connectors across ChatGPT Enterprise and Claude Enterprise with encryption boundaries, auditability, and least-privilege access for a large enterprise.
- Headline number
- 2 platforms enabled
- Role
- Built
- Category
- Security & Operations
- Updated
Architecture
- ChatGPT EnterpriseClaude Enterprise
- Oktaexternal OAuth
- Least-privilege roles
- Snowflake datacustomer-managed keysAudit evidence
Key takeaways
- Enterprise key management validated in a sandbox before any production data moved
- Custom and external OAuth in Snowflake with Okta as the identity provider
- Audit evidence captured for security review, so approval took one meeting
- Both platforms enabled on the same access model, so policy lives in one place
The situation
Leadership wanted enterprise AI assistants connected to company data. Security wanted proof that keys stayed under company control, access followed existing identity, and every call could be audited. Both were right.
What I built
I stood up a sandbox to validate the enterprise key management path end to end, then designed the Snowflake access model: custom OAuth for first-party apps and external OAuth through Okta for the assistant platforms, with roles that grant the minimum each connector needs. I captured the audit evidence as I went so the security review had artifacts instead of promises.
The outcome
Both assistant platforms went live against governed data with the same access model. The security team got a repeatable pattern for the next connector instead of a one-off exception.
Stack
Related work
Same discipline, different system.
Need this built inside your organization?
One person, architecture to deployment. Tell me what is stuck and I will tell you where I would start.