SLO discovery Labs
SLO discovery is a conversational assistant that helps you decide what to measure before you build an SLO. You describe a service in plain language, the assistant interviews you with follow-up questions, and it turns the conversation into draft SLO definitions you can review and hand off.
It helps start the SLO conversation. It does not create SLOs for you.
SLO discovery is a Nobl9 Labs feature. It can be unavailable, behavior can change without notice, and sessions can be lost during iteration. Generated output can be incomplete or wrong. Don't rely on it for regulated, contractual, or auditable SLO design.
What SLO discovery does
SLO discovery interviews you about one service and proposes draft SLO ideas based on what you tell it. Each proposal describes what to measure, why it matters, and what still needs clarification.
Use it to:
- Think through the first SLOs for a service.
- Prepare for an SLO design conversation with your team.
- Onboard people who are new to SLOs.
SLO discovery is not:
- Automatic SLO creation. Nothing is created in your organization.
- Production-ready SLO authoring. Output is draft guidance, not applicable Nobl9 YAML.
- A replacement for an SRE or solutions engineer. A person still reviews the proposals and builds the real SLOs.
Availability
SLO discovery is a Labs feature enabled per organization. If you don't see the entry point described below, contact Nobl9 support to have it turned on for your organization.
Opening SLO discovery
On the Service Level Objectives page, click Create SLO and select Discovery session from the menu. The option carries a Labs badge.
Running a discovery session
Describe your service
Start a new session and describe the service you want SLOs for. Keep the first message short and concrete — name the service and its main job. For example:
The checkout API. It processes payments for our e-commerce customers, and the key journey is completing a payment and receiving confirmation.
A single service with one clear user journey works best. Avoid sensitive data — treat a session as a working draft, not a system of record.
Answer the interviewer's questions
The assistant asks follow-up questions to fill in what it needs: the scope of the system, the critical user journeys, how the service can fail, and what acceptable quality looks like. Answer in your own words, and add detail wherever the assistant's assumptions look wrong.
A topic panel tracks how much of the discovery has been covered, grouped into core topics and contextual topics. You don't have to complete every topic — it's there to show where the conversation still has gaps.
Generate SLO definitions
Once the assistant has enough context, it tells you so and the Generate SLO definitions button becomes available. You can generate at that point, or keep the conversation going to add more detail first. If later answers make the picture less clear, the assistant can ask for more before it's ready again.
Click Generate SLO definitions when you're ready to turn the conversation into proposals.
Reviewing SLO proposals
Generated proposals appear in the SLO proposals panel. Each proposal is a structured draft that typically describes:
- The user journey in scope and the boundaries of what the SLO covers.
- A recommended service level indicator and what a successful state looks like.
- A suggested success target and the rationale behind it.
- The risks of failure and how failure would affect the error budget.
- Points to calibrate once you have real data.
Treat every proposal as a starting point. A person reviews it, adjusts it, and uses it to create the actual SLO in Nobl9 — proposals are guidance, not configuration, and can't be applied directly.
Managing sessions
Past sessions are listed so you can return to earlier discovery work. You can search the list by name, rename a session to something memorable, and delete sessions you no longer need.
Deleting a session is permanent — it removes the chat history and any SLO proposals generated in it.
Sending feedback
SLO discovery is still being trained, and feedback directly shapes it. You can rate individual assistant responses, and give feedback on the session as a whole when you finish. Helpful feedback includes bad or repeated questions, wrong assumptions, weak proposals, and missing context.
Limitations
As a Labs feature, SLO discovery has limits you should plan around:
- Output is draft guidance. Proposals aren't Nobl9 YAML and can't be applied to create SLOs directly.
- Sessions aren't durable. Chat history, proposals, and feedback can be deleted, lost, or changed during Labs iteration. Copy anything you want to keep.
- Behavior can change. The interface, topics, and generated output may change without notice.
