Summary
On many internal platforms, getting a service still means filing a request that an operator reads before triggering the pipeline that provides it. The article draws on guidelines the author worked on for moving such a platform to a self-service catalogue. Placed on an effort-versus-impact chart, the step from the ticket-driven platform to self-service is short compared with the climb from legacy to the platform, and the remaining effort falls on three subjects: wiring an interface to the existing pipelines, watching how the services are used, and training. Only one of them is technical. In the impact grid, tools and culture both score 3 out of 4, but a tool has a delivery date and a professional value does not.
A Kafka case makes it concrete. A central team of five full-time equivalents ran the message bus for 150 applications and 15 000 topics, and topic requests came through tickets until their volume became unmanageable. The ticket queue was replaced by one YAML file per application in a Git repository, a pipeline that validates every change against predefined rules and applies it outside production, and a tag that the central team creates for production. Deleting a topic stayed manual. The results are qualitative: no more tickets to create or modify a topic, autonomous teams, requests handled in parallel rather than in sequence. The source gives no lead-time figure, and the article does not invent one.
The leadership argument is that the tooling only encodes a decision about where trust lies. The central team stopped executing requests and kept two things: the rules and the production gate. Trust is earned through training badges, usage ambassadors, metrics and alerts rather than by locking the interface. The platform team stops being the intermediary and becomes the owner of eligibility criteria, terms of use, indicators and the catalogue as a product. Without these decisions, automation only moves the queue.
Key ideas
- The step from a ticket-driven platform to self-service costs far less effort than the move from legacy to the platform, and that effort is mostly cultural.
- Tools and culture weigh the same in the impact grid, but only the tool has a delivery date; the levers are training, communication and sponsorship, not automation.
- Topics as code replaced a ticket queue with one YAML file per application, a validating pipeline and a production tag, while deletion, an occasional task, stayed manual.
- Self-service is a decision about where trust lies: the central team keeps the rules and the production gate, and trust is earned through training and usage ambassadors.
- The platform team stops executing requests and starts owning eligibility criteria, terms of use, indicators and the catalogue as a product.