Verify: Pipeline Agents - FY27
**Epic Owner:** TBD (Product Manager - Verify Stage)\ **Engineering Owner:** TBD (Engineering Manager - Pipeline Execution or CI Platform)\ **FY27 Company Objective:** AI Modernization Journey\ **FY27 CTO Program:** Duo Agent Platform Adoption\ **Key Result:** DevOps agents: 140K LLM requests from DevOps first-party agents --- ## Overview Launch a suite of AI-powered **capabilities** that transform how users interact with GitLab CI/CD pipelines, starting with pipeline authoring assistance and expanding into pipeline failure remediation, optimization, observability, and configuration management. ## Business Objective Address GitLab's CI adoption gap by providing intelligent assistance that guides users from pipeline creation through advanced configuration, optimization at scale, reducing time-to-first-pipeline and improving pipeline efficiency for existing users. ## Agent Portfolio - Use Cases Prioritized for Delivery | Ordinal Priority Ranking | Pipeline Lifecycle | Pipeline Agent Use Cases | Target Delivery | Problem to Solve | Notes | |--------------------------|--------------------|--------------------------|-----------------|------------------|-------| | 1 | Authoring | https://gitlab.com/groups/gitlab-org/-/work_items/21405+s | Q2 - 19.2 | | | | 2 | Execution | https://gitlab.com/groups/gitlab-org/-/work_items/21181+s | Q2 - 19.1 | Trigger the flow automatically on GitLab Internal projects only | | | 3 | Execution | https://gitlab.com/groups/gitlab-org/-/work_items/21837+s | Q2 - 19.1 | Fix suggestions in same MR and improved flow outcomes (higher percentage of fixes) | | | 3 | Execution | https://gitlab.com/groups/gitlab-org/-/work_items/21300+s | Q2 - 19.1 | Optimize Credit Consumption for Fix Failing Pipeline Flow | | | 4 | Execution | [Fix a failing pipeline - auto trigger the flow (Customer projects)](https://gitlab.com/groups/gitlab-org/-/work_items/21225) | Q2 - Tentative 19.3 | Trigger the flow automatically on customer projects | | | 5 | Execution | https://gitlab.com/groups/gitlab-org/-/work_items/20921+s | Q3 | Detect and retry failing pipelines on flaky test failures | Required: Building a Flaky Test Detection feature | | 6 | Execution | https://gitlab.com/groups/gitlab-org/-/work_items/20623+s | Q4 | Detect and retry failing pipelines on infrastructure failures | | ## ## Success Metrics / Business Outcomes | Metric | What? | Current | Target (FY27Q4) | |--------|-------|---------|-----------------| | Average time to fix a failing pipeline (all GitLab.com projects) | Track customer outcome for Fix Failing Pipeline Flow | | | | CI Expert Agent Adoption | Track CI Expert Agent invocations | | | | Fix Failing Pipeline Flow Adoption | Track Monthly Sessions/Users | 1.6 (Apr 2026) | 4 | ## Important Reference Links - Please Read if new to Duo Development #### AI links - [Building with Duo Agent Platform](https://docs.google.com/presentation/d/1RLj4LIt8yPnjN5OWYAsP4tSNqQJMXJ_e819yHUdE0us/edit?slide=id.g375873d1863_0_0#slide=id.g375873d1863_0_0) - [Flow Registry Framework Developer Guide](https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/ai-assist/-/tree/main/docs/flow_registry) - [Tools](https://gitlab.com/groups/gitlab-org/ai-powered/-/wikis/home/Agentic-IDE#tools-agenticidetool) - [Interaction methods epic](https://gitlab.com/groups/gitlab-org/-/epics/18712) - [Agents & flows](https://gitlab.com/groups/gitlab-org/-/epics/19001) - [DAP tracker](https://gitlab.com/groups/gitlab-org/ai-powered/-/wikis/home/Program-Duo-Agent-Platform/DAP-Enablement/) - [Duo Agent Platform Features](https://gitlab.com/gitlab-org/ai-powered/duo-agent-platform/-/wikis/product-engineering/platform-roadmap/Duo-Agent-Platform-Features) - [Duo Agent Platform Docs](https://docs.gitlab.com/development/ai_features/duo_agent_platform/) - [Instrumentation](https://gitlab.com/groups/gitlab-org/-/epics/19545) -[Deep Research Agent Prototype](https://gitlab-research-agent.michaelangelo.io/) - Knowledge Graph - https://gitlab.com/gitlab-org/rust/knowledge-graph - https://gitlab-org.gitlab.io/rust/knowledge-graph/
epic