How do we keep this defensible to regulators?
Governance is in the architecture: observability, audit trails, policy enforcement, escalation paths, model registries. We co-design the controls with your risk and legal teams during the Architecture Engagement. The output is something an internal auditor or external regulator can trace.
What happens when the next frontier model releases?
Models are swappable. The orchestration layer abstracts the model boundary so a migration is a configuration change, not a re-architecture. Continuous Partnership covers the migration cadence; without it, your team handles it using the documented procedure.
Can we run this without sending data to OpenAI / Anthropic / etc.?
Yes. Local Inference, Hybrid, and Private Enterprise patterns are designed for exactly that. We pick the pattern based on your data classification, regulatory environment, and capability ceiling: not based on which vendor we like.
How does this differ from a Big-Four transformation programme?
We deliver working systems and transfer ownership. We don't sell open-ended advisory hours, we don't run perpetual programmes, and we don't recommend a vendor we then resell.
What's the realistic minimum we should spend?
If you need a defensible plan only, R180k buys a Strategic Diagnostic. Anything below that, you're better off with Build Discovery. Anything above, the engagement model is sized to your function and risk profile.