After several brittle point-to-point chatbot integrations fail, the city adopts a platform integration standard for AI tools. What ecosystem practice is this?

Standardizing how AI tools integrate across municipal systems after point-to-point failures is the correct ecosystem practice. Brittle one-off chatbot connections multiply credential patterns, logging gaps, and breakages whenever a source system changes. A published integration standard—gateway use, auth patterns, error handling, data classes, and observability—lets new AI tools connect predictably and lets platform teams support them without reinventing plumbing for every department. The practice works because municipal AI ecosystems need reusable, governable integration patterns once point-to-point sprawl proves fragile.

Encouraging every department to invent its own one-off integration style fails by concept: more unique styles recreate the brittle point-to-point failures that triggered the change. Documenting nothing so future vendors can guess how to connect fails by concept—undocumented expectations produce insecure improvisation and inconsistent implementations. Forbidding all integrations so chatbots stay isolated forever fails by concept: isolation avoids some risk but also blocks the systems-of-record value that makes civic AI useful for permitting, 311, and casework.

Exam caveat: standardization follows demonstrated failure of unmanaged point-to-point links; it is an ecosystem maturity step, not a ban on integration. Keep municipal multi-system framing. Operational check: publish an AI integration standard (API gateway, SSO, logging, data handling); require new tools to conform; migrate the worst point-to-point links; and measure reduced break/fix and faster onboarding for subsequent AI connectors.