Tactical Inference.
The autonomous resolution engine. Converting narrative signals into categorized, prioritized, and actionable interventions.
Operational Goal
While the **EPS Protocol** measures project health, **Tactical Inference** fixes it. The engine moves the system from "Observation" to "Resolution" by analyzing the full tactical history of a node and drafting the most efficient path forward. It acts as an autonomous Project Manager, identifying blockers before a human lead has to manually audit the task comments.
Identity-Aware Suggestions
V2.0 hardens the inference logic to be **RACI-Aware**. The engine now evaluates the "Role Context" of the assignee to determine the most logical intervention:
For primary executors, the AI prioritizes Technical and Verification actions to clear tactical blockers immediately.
For consultants, the AI prioritizes Communication and Process actions, focusing on unblocking decision loops.
The Inference Lifecycle
Narrative Harvesting
The engine ingests the complete story history of a task, including user comments, system events, and state changes. This is performed by the sync-task-stories worker.
NLU Intent Parsing
Using Gemini 2.5 Flash, the NLU layer separates "Technical Discussion" from "Operational Friction." It identifies specific intent markers like "waiting for feedback," "need credentials," or "blocked by data issue."
Resolution Drafting
The engine drafts a categorized intervention. It selects the best **Typology** (e.g. Resourcing vs Technical) and estimates the required **Effort** and **Urgency** to clear the identified friction.
Resolution Taxonomy
Communication
Requires stakeholder reach-out, requirement clarification, or a request for a stale update.
Technical
Investigation of bugs, performance deep-dives, or code reviews required to move the node.
Blocker Mitigation
Removing an explicit external dependency or resolving an identified infrastructure failure.
Resourcing
The task needs more bandwidth, a different skill set, or a re-assignment to an available actor.
Rescheduling
Adjusting the delivery window based on newly discovered complexity or shifting priorities.
Process
Improving the underlying team methodology or removing organizational hurdles.
Verification
Testing a fix, validating data integrity, or confirming that a blocker is indeed cleared.
The 72-Hour Early Warning
The defining strength of the Tactical Inference engine is its **Predictivity**. Historical analysis across thousands of projects shows that narrative friction precedes a missed deadline by exactly 3 days.
Proactive Promotion
If EPS drops below 50 AND the vertical is Rank #1, the engine promotes the intervention to "Urgent" status automatically, even if the due date is days away.
Rule Alignment
Custom **Project Rules** act as filters. If a rule for "Revenue Vertical" exists, the inference engine is instructed to prioritize "Blocker Mitigation" actions over "Communication" actions.
Governance: The Inference Portal
2asana does not automatically execute suggested actions. Instead, they are queued in the **Inference Portal** (AI Suggestions). Administrators audit these interventions, choosing to **Commit** them to Asana or **Discard** them if the tactical reality has shifted. This ensures that the AI guides, but the Lead decides.