Build a multi-tenant SaaS with AI by making tenant identity and isolation part of every contract from the first schema. Define what a tenant is, how users join it, which roles exist, how data is partitioned, and which operations may cross tenant boundaries. Then implement one journey end to end.
Model tenant ownership explicitly in data, requests, jobs, files, caches, logs, and external integrations. Enforce authorization on the server and at the data boundary rather than relying on hidden interface controls. Ask the agent to generate tests with two tenants and several roles, proving both allowed access and forbidden cross-tenant reads and writes. Cover invitations, role changes, account removal, exports, background jobs, retries, and support access. Keep billing entitlements separate from authorization unless the product contract deliberately connects them. Review migrations for existing tenant data and define recovery before applying them.
Use shared or isolated infrastructure based on risk and operating needs, not a generic prompt. Record the chosen isolation model and its tradeoffs. Add tenant-aware observability without leaking one customer’s data into another customer’s logs or alerts.
Start with one small tenant-scoped resource and a negative isolation test. Require human security review before production. AI can scaffold repeated patterns and tests, but the team must own the tenancy model. Expand only after each boundary is explicit, enforced, observable, and recoverable.
