DraftThis concept page is in progress. The structure is in place, individual exhibits are still being designed. The agent-native cycle is the polished reference.
Hybrid Mode, Designed
Hybrid is the configuration most teams actually adopt first. Add the agent team as a peer to your existing human dev team. Both commit to the same sprint. Start the agent team on simpler, well-scoped stories where confidence is easy to build. Migrate more complex work over as the trust grows. The team stays in control of where the boundary moves.
The walkthrough below uses the same scenario as the agent-native cycle and the sprint-mode cycle, routed through both teams this time.
The Scenario
Same scenario as the other concepts: the CS team needs an account health widget on the Account Detail page. In hybrid mode, the work routes through a sprint that contains both a human dev team and an agent team. The agent team gets the well-scoped widget. The human team takes a parallel Account Detail layout redesign that needs more judgment. Mid-sprint, scope expands and the agent team becomes a flex-capacity buffer.
ProductIntel · Work · Backlog · Sprint 47 candidates
AI-suggested team tags
Account health score widget on Account Detail
Well-scoped UI + service · existing patterns
Redesign Account Detail layout for density
Design judgment · customer feedback needed
CSV export for case list
Standard pattern · API + button + download
New onboarding tour for support reps
Content + UX flow · stakeholder review
Filter by tag in case list
Could go either way · simple but no urgency
What humans do
Drafts the stories. Reviews team-tag suggestions. Overrides if the AI guessed wrong.
What agents do
Suggests team tag (human / agent / either) per story based on scope, code patterns, and risk. Surfaces any story it cannot confidently classify.
ProductIntel · Work · Sprint Planning · 47
Human team
32 pts
5 devs · default
Agent team
8 pts
conservative · sprint 3 with agents
Total committed
38 pts
2 pts unallocated buffer
Pulled into Sprint 47
AI suggestion
Agent team capacity is intentionally conservative for sprint 3. Both pulled stories follow patterns the agent team has shipped before (widget, CSV export). Recommend reviewing for capacity expansion in Sprint 48 retro.
What humans do
Confirms team assignments. Calibrates how much agent capacity to allocate this sprint based on confidence.
What agents do
Suggests sprint pull for both teams. Recommends the agent team start with simpler, well-scoped stories early. Surfaces any cross-team dependencies.
ProductIntel · Work · Board · Sprint 47 · Day 5
Backlog
14In Progress
3Review
2Done
4ProductIntel · Work · Sprint Risk · AI Detected
Account Detail redesign (13 pts, human team) is blocked on customer feedback that is now scheduled for next Tuesday. Sprint commitment of 38 pts is unlikely to land. Recommend reallocating two unstarted backlog stories to the agent team to keep the sprint on track.
Proposed flex move
CSV export for case list
Backlog (this sprint) → Agent team · in progress
Bulk delete for archived accounts
Backlog (this sprint) → Agent team · in progress
Sprint commitment
25 of 38 pts on track. +5 from flex = 30 of 38. Account Detail (13 pts) carries to Sprint 48 with customer feedback in hand.
Why these two
Both follow patterns the agent team has shipped. No cross-team dependencies. Estimated agent runtime: under 2 hrs combined.
What humans do
Manages the sprint. Approves any flex-capacity reallocations the AI proposes. Reviews PRs from both teams.
What agents do
Tracks burn rate, blockers, and scope changes. When sprint commitments are at risk, proposes shifting unstarted work to the agent team to keep the sprint on track.
ProductIntel · Work · Per-Story Execution Mode
Stories assigned to human team follow the standard development workflow. AI assists with estimation, triage, and blocker detection.
Stories assigned to agent team run the agent-native mini-cycle. PM stays in the loop at Resolve and Review.
Same PR queue
Both flows end in PRs reviewed and merged by humans. The human team sees both teams’ work in the same review queue. Consistent quality bar across the sprint.
What humans do
Human team uses traditional flow: design, code, review. Reviews all PRs from both teams.
What agents do
Agent team uses the agent-native mini-cycle (refine, resolve, execute) per story. Spawns isolated worktrees, commits, opens PRs.
ProductIntel · Work · Sprint 47 · Closeout
Planned
38 pts
includes flex
Delivered
35 pts
92% of plan
Carried
13 pts
Account Detail to Sprint 48
Flex used
5 pts
agent team buffer
Human team
19 of 32 pts delivered. Onboarding tour and KB search shipped. Settings reorg shipped. Account Detail blocked on customer feedback, carried.
Agent team
13 of 13 pts delivered (8 planned + 5 flex). Account health widget, CSV export, bulk delete all shipped without follow-up issues.
AI summary
Sprint 47 delivered 35 of 38 planned points. The flex move on day 5 added 5 points to the agent team and prevented the sprint commitment from missing badly. Account Detail carries to Sprint 48 with customer feedback now scheduled. Agent team delivered everything they pulled, including flex, with zero rework.
What humans do
Demos to stakeholders. Marks sprint complete. Reviews per-team velocity, scope creep, and flex-capacity usage.
What agents do
Generates the sprint report. Distinguishes human-team and agent-team contributions. Flags whether the flex move kept commitments on track.
ProductIntel · Work · Sprint 47 · Retro
What worked
What to change
AI noticed · boundary suggestion
The agent team has delivered well-scoped frontend + service stories cleanly across three sprints. Two stories in the current backlog (filter by tag, archive bulk action) match the same pattern. Recommend assigning to the agent team in Sprint 48 to grow default capacity from 8 to 12 pts.
What humans do
Facilitates the retro. Captures what worked, what to change. Decides whether to grow the agent team's default share for next sprint.
What agents do
Surfaces patterns: which story shapes the agent team handled cleanly, where the flex move helped, and which currently human-only work could be tested in the agent stream next sprint.
One Sprint, Two Teams
Hybrid is not a hedge. It is the actual adoption path for a team that already runs sprints. The agent team starts as a small share of capacity, takes on the well-scoped work where confidence is easy to build, and earns more share over time. When the sprint hits trouble, the agent team is also the flex capacity that keeps commitments on track.
The team stays in control of where the boundary moves. The platform does not push the change.