Dependencies are the invisible schedule

Kanban boards show what is in flight. They rarely show what must finish before what unless teams discipline themselves to record blockers. When dependencies live only in standup speech, planners discover the critical path during a demo that cannot ship—because integration was waiting on a API task nobody linked.

AI task dependency analysis does not replace your PM's judgment about sequencing. It surfaces relationships already implied in tasks—explicit links, blocker comments, approval gates, cross-team handoffs—and helps you ask better questions before dates commit.

WKFGo models dependencies through add_dependency and related task tools, with AI assistants reading the graph via MCP and chatbot queries grounded in live data.

Problem: three dependency blind spots

Undocumented blockers — "Waiting on backend" in a comment is not a dependency edge. Search cannot traverse it; reports cannot age it.

Cross-project assumptions — Team A assumes Team B's milestone is done because the slide was green. No link exists in either project.

Dependency sprawl — Fifty tasks with pairwise links become unreadable. Teams need views that highlight chains touching a release or a person—not a hairball diagram for its own sake.

AI helps by summarizing chains and proposing links for human confirmation. It should not silently create dependencies without an owner approving.

Framework: record → traverse → analyze → act

Record explicitly — Use add_dependency when task B truly cannot start until task A completes (or specify the relationship your process uses)
Traverse from anchors — Start from release tasks, milestone columns, or integration epics; walk upstream
Analyze with AI — Prompt: "List dependency chains blocking Release 2.4; include owners and ages"
Act on bottlenecks — Reassign, descope, parallelize with documented risk, or escalate to decision_inbox

The graph is only as honest as team habits. Analysis amplifies good data; it cannot fix silent blockers.

Signals worth linking

Hard technical prerequisites (schema migration before feature)
Approval gates (legal review before customer comms)
External vendor deliverables tracked as tasks
Cross-project integration points with named counterpart tasks

Signals that should stay comments

Soft preferences ("nice if design lands first")
Morale blockers needing human conversation
Speculative future work

Solution: dependency analysis in WKFGo

Weekly release readiness pass

Before committing a date:

Pull release tasks via list_release_tasks or label filter
For each, ask AI/MCP to enumerate upstream dependencies with status and assignee
Flag any chain where upstream is In Progress past due or unassigned
Add missing add_dependency links discovered in the meeting—do not leave them verbal

Cross-project portfolio view

For programs spanning multiple projects, run smart_search or portfolio queries for integration keywords, then validate links exist both directions. AI can suggest candidate pairs; PM confirms.

Pair with flow aging

flow_aging and control views show work stuck in columns. Dependency analysis explains why aging happens—is it blocked upstream or starved of review capacity? Combine both lenses in sprint review.

Escalation path

When a chain threatens a contractual date, add_decision_item or monitor-driven decision_inbox entries force an executive choice: delay, descope, or add capacity—with simulate_scenario comparing paths.

AI prompt patterns (examples)

"Which tasks block TASK-1234 directly and transitively? List owners."
"Find tasks in Project X with 'blocked by' in comments but no dependency link."
"Summarize dependency depth for sprint goal tasks; highlight depth > 3."

Always require links back to task IDs in outputs so humans verify quickly.

Anti-patterns

Linking everything — dependency graphs become noise
AI auto-linking without confirmation — creates false critical paths
Ignoring approval dependencies — legal gates are dependencies too
Analysis without action — produce a weekly "top five chains" SLA for resolution

Dependency chains break at integration boundaries. Map blockers with add_dependency early; weekly ask MCP for tasks blocked by a single upstream owner. AI highlights critical paths from live links—not inferred relationships from task titles alone.

Run a monthly audit: do AI answers cite task IDs from MCP? If not, tighten prompts to require citations before any stakeholder-facing draft leaves the team.

Permission-aware tools mean the same question gets different grounded answers for PM vs contractor—that is a feature, not a bug. Train teams to expect scoped truth.

FAQ

Does AI auto-schedule my project?
No. It analyzes and summarizes relationships; humans change assignments and scope.

add_dependency vs comments?
Comments narrate; dependency edges power traversal, reports, and AI queries.

Cross-project dependencies?
Supported when tasks are linked across projects; confirm permissions expose both sides to analysts.

How is this different from Gantt?
Gantt visualizes dates; dependency analysis validates logical prerequisites behind those dates.

Map blockers before they map your slip

Record dependencies honestly, analyze chains before commit dates, and escalate forks with evidence.

Try it now

Put these patterns on live project data—not slide decks.