Engineering Task

Engineering tasks exist only in developers' heads or scattered across sticky notes and personal to-do lists. 'What are you working on?' gets a different answer depending on when you ask. There is no shared task tracker, no backlog, and no way to see what the team is building or how far along any work item is.

None — AI cannot forecast velocity, balance workload, or plan sprints because no engineering task records exist in any system.

Engineering tasks are tracked in a tool like Jira or Linear, but practices are inconsistent. Some tasks have detailed descriptions; others have a two-word title and no context. Story point estimates exist on some tasks but not others. Task types are mixed — bugs, features, and tech debt all use the same generic 'task' type. 'Can you estimate when this feature will ship?' requires manually reading through every ticket.

AI can list open engineering tasks and filter by assignee, but cannot forecast completion, estimate velocity, or identify blockers because task metadata is too sparse and inconsistent to support analysis.

Engineering tasks follow a defined schema with required fields — task type, story points, acceptance criteria, priority, and linked epic. Sprint planning uses consistent velocity metrics. Task states follow a defined workflow (To Do, In Progress, In Review, Done). But engineering tasks are standalone documents — they don't link to the PRs that implement them, the deployments that ship them, or the production metrics that validate them.

AI can forecast sprint velocity, identify at-risk engineering tasks based on time-in-status, and suggest workload balancing across the team. Cannot trace task completion through to production impact because tasks don't connect to downstream delivery artifacts.

Engineering tasks are comprehensive work records with full delivery context. Each task links to implementing PRs, deployment records, and production validation metrics. A product manager can query 'show me all engineering tasks for Feature X, their implementation status, which have shipped to production, and whether customer satisfaction scores changed after delivery' and get a complete answer.

AI can provide end-to-end delivery intelligence — from task creation through implementation, deployment, and business impact. Can predict task completion dates based on historical patterns for similar task types. Cannot yet auto-decompose large tasks because decomposition criteria aren't formalized.

Engineering tasks are formal entities in a delivery ontology with machine-readable decomposition rules, typed relationships to all delivery artifacts, and quantified completion criteria. An AI agent can take a product requirement, decompose it into engineering tasks following established patterns, estimate each task based on historical velocity for similar work, and assign based on team capacity and expertise.

AI can autonomously manage the engineering task lifecycle — creating, estimating, assigning, and tracking tasks within the formal delivery model. Human judgment is needed for novel work that doesn't match established patterns and for strategic prioritization.

Engineering tasks are self-maintaining entities that auto-adjust in real-time. Estimates recalibrate as the team's actual velocity evolves. Priorities reorder when production incidents or customer escalations shift business urgency. Dependencies update when blocking tasks complete or slip. The engineering task backlog is a living system that reflects current reality without manual grooming.

Fully autonomous engineering task intelligence. AI creates, estimates, prioritizes, assigns, and adjusts engineering tasks in real-time based on team capacity, business priorities, and production signals.

Ceiling of the CMC framework for this dimension.