The client needed access, not a database replacement
The client's operational database already supported the application and existing automations. Replacing it would have added risk, migration work, and cost without solving the narrower problem: giving approved people and AI tools controlled access to specific information.
I separated those responsibilities. The existing database remained the only place where operational records were created and changed. AWS received a private, read-only copy containing only the approved business data needed for employee and AI access.
That decision avoided a disruptive migration and prevented the new access layer from becoming a second source of truth. The access layer was designed to let approved users retrieve the information they needed without receiving production database credentials or unrestricted query access.
A ~$7,000 annual quote became a much smaller running-cost estimate
The quoted access option was roughly $7,000 per year. The alternative I built added an estimated $38–$45 per month in AWS service costs, or $456–$540 annually. That leaves a modeled difference of $6,460–$6,544 a year, rounded to about $6,500.
The business benefit was a lower-cost route to the access the client needed, while preserving the system they had already invested in. A broader managed option can make sense when its included services justify the price; here, the narrower requirement made a focused build worth considering.
Lower cost without disrupting the working application
I kept the client's database and automations in place. The new access layer holds a read-only copy of approved information, designed for employees and AI tools without giving them credentials to the production database.
Information refreshes every 15 minutes, and each update is checked before it goes live. If an update fails, the last good copy stays available. That keeps a failed refresh from replacing usable information, although the copy can be older until a later refresh succeeds.
The result: a ~$6,500 annual cost difference in the model
The client had a focused alternative to the roughly $7,000 annual option, with a much smaller estimated infrastructure bill and no need to replace the working database.
- Roughly $6,500 a year between the quoted option and the estimated added AWS cost.
- No database migration or replacement of existing automations.
- A controlled route to approved information, with access rules tested before broader rollout.
This is a modeled operating-cost comparison, not a measured reduction in an existing bill or a completed return-on-investment calculation.
How the estimate was calculated
The ~$7,000 annual figure was a client-specific quote, not a public Supabase price. The AWS estimate was $38–$45 per month: $7,000 minus 12 months of those costs gives $6,460–$6,544, rounded to ~$6,500 per year.
The estimate excludes development, maintenance, the client's existing Supabase and Claude subscriptions, and possible Supabase data-transfer charges. Those costs must be included in a full buying decision. Actual AWS costs depend on usage; employee rollout and realized savings were not established by the service-level tests.