Overdue tells you you are late—aging tells you why

Project delay prediction in mature teams is not a mystical algorithm that prints a slip date. It is early detection of flow dysfunction plus explicit scenario modeling before executives pick a fix. By the time tasks turn overdue, options narrow to apology, crunch, or scope cut without analysis.

WKFGo combines two honest MCP capabilities for delay intelligence:

• flow_aging — Surfaces work that has sat too long in a column, often before due dates break
• simulate_scenario — Projects trade-offs for modes like delay_release, cut_scope, add_people, and freeze_project

AI assistants connect these tools to natural-language briefings—always citing returned data, never inventing percentages.

What actually predicts delay

None of these alone is "AI prediction." Together they form a leading indicator dashboard your assistant can narrate weekly.

flow_aging: stuck work before red dates

Flow aging measures how long work stays in its current state—often mapped to Kanban columns. A task due next week that has been In Review for twelve days is already a delay risk even if the calendar says green.

Weekly control ritual (15 minutes):

1. Pull oldest items per column via **flow_aging**2. For each: blocker? owner? split? escalate?
2. One systemic fix—WIP limit, approval SLA—not only firefighting

Pair aging standup questions: "What has been In Progress more than five days?" and "Who unblocks it by when?"

simulate_scenario: decision forks with stated assumptions

When delay is likely, leadership needs options, not panic. simulate_scenario models strategic forks across dimensions with confidence and assumptions documented—so the debate is explicit.

Example prompts via MCP:

• "Simulate delay_release two weeks on Project Atlas—what milestones shift?"
• "Compare cut_scope on Feature X vs add_people to integration squad."
• "Model freeze_project on legacy migration while we stabilize core."

Simulation supplements judgment; it does not auto-execute changes. Humans resolve via decision_inbox and log outcomes for calibration.

AI's role: narrate, don't hallucinate

Connect Cursor or Claude to WKFGo MCP:

1. "Run flow_aging for Project Orion—top ten oldest in Progress and Review."
2. "Summarize overdue and blocked tasks with IDs."
3. "Given aging hotspots, which simulate_scenario mode fits—delay, descope, or staff?"

Require citations: task ID, column, age days, scenario assumptions. If the model cannot cite tool output, treat the answer as draft fiction.

Anti-patterns

• Waiting for overdue reports only
• Treating simulation output as prophecy without reading assumptions
• Running scenarios once in chat, never logging the decision
• Ignoring Review/Test columns where real bottlenecks hide

Connecting delay prediction to delivery rhythm

• Daily — Aging items in standup
• Weekly — flow_aging control meeting + one scenario if slip trend worsens
• Steering — executive_brief + finance_summary when delay threatens budget or client date
• After decision — decision_inbox resolution + decision log for next cycle

Leading vs lagging indicators

Treat overdue counts as lagging—they confirm failure. Treat aging percentiles, WIP age, and approval queue depth as leading—they show where flow will break next week if nothing changes. AI assistants excel at comparing this week's aging top ten to last week's and narrating trend direction; humans decide whether the trend warrants simulate_scenario or a scope conversation with the client.

Document one "delay early warning" threshold per project—e.g., any item >7 days In Review triggers facilitator action. Thresholds differ by team; the point is consistency, not a universal magic number.

Stakeholder communication without fiction

When delay is likely, PMs owe stakeholders options and dates tied to tasks—not reassurance. MCP workflow: pull overdue and aging lists with IDs, draft status email citing those IDs, human edits tone and sends. get_project_report can supplement narrative for steering audiences. Never send AI-drafted delay notices without verifying task states in the UI once while building team trust.

FAQ — AI delay prediction

Does WKFGo auto-predict a slip date?
It exposes signals (flow_aging, reports) and scenario tools—not a single magic forecast field.

Which simulate_scenario mode first?
Start with the fork leadership is actually debating—usually delay_release vs cut_scope.

Can AI replace the PM on delay communication?
AI drafts status from live tasks; humans own stakeholder messaging and commitments.

How is this different from Gantt critical path?
Gantt shows plan; aging shows reality in flow. Use both.

Try it now

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

• Create your free account
• Tour the full feature set
• See plans
• MCP documentation
• Contact us