Why Enterprise AI May Start with Dedicated Customer Instances

Editorial placeholder -- not yet a full draft.

This article will convert the internal architecture decision record (docs/architecture-decisions/ADR-0001-dedicated-instance-first-deployment.md) into a vendor-neutral discussion for a public audience.

Planned scope:

• Why AI products that store governance evidence, assessments, and organisational context carry higher isolation stakes than typical SaaS data.
• The trade-off between shared multi-tenant SaaS (lower cost, faster onboarding, but a single RLS/authorization defect can cross customers) and dedicated-instance deployment.
• Why a dedicated-instance-first approach does not have to mean customer-specific forks -- one versioned product, multiple deployment modes.
• Preserving a credible future path to shared SaaS without adopting its risk profile before the product needs it.

Publish only as a general industry discussion -- this must not read as, or reference, an internal decision record, and should not name specific internal deployment mechanics.