Insights
Practical notes on AI delivery, SaaS architecture and modernizing critical systems.
Agentic AI
Start with the consequence of an action, not the capability of the model. A read-only lookup and a change to a customer account need different levels of authority.
Read the perspectiveModernization
Begin by understanding what must keep working. Business rules, downstream integrations, data semantics and recovery requirements shape a safer modernization sequence.
Read the perspectiveSaaS architecture
Growth exposes the boundaries you did not define: tenants, permissions, configuration, entitlements, data ownership and operating responsibilities.
Read the perspectiveAI delivery
Production AI applications need approved data, permissions, evaluation, cost visibility and clear recovery paths.
Read the perspectiveAI evaluation
Accuracy alone hides the operating truth. What matters is whether people complete the task, trust the result and recover when it is wrong.
Read the perspectiveDelivery risk
A rollback plan is a design artifact. If it is written after the migration starts, it is usually a wish rather than a plan.
Read the perspectiveModernization
A rewrite is justified when the current design cannot support the business rules you must keep. It is a poor answer when the problem is only delivery pace.
Read the perspectiveSaaS architecture
Multi-tenancy is a commercial decision before it is a technical one. Pooled, dedicated and hybrid models carry different operating costs, isolation guarantees and support responsibilities.
Read the perspectiveProduct decisions
Not every boundary matters equally. Rank them by what breaks first: permissions, configuration, entitlements or data ownership.
Read the perspective