Assignment Algorithm

Munera's assignment engine combines GPT-4 skill matching with multi-factor scoring to rank engineers for each task. The result is a prioritised list of candidates with confidence scores and plain-English explanations.

Scoring factors

When a task enters pending_assignment status, the engine evaluates every active engineer on five weighted factors:

FactorWeightDescription
Skill match40%Overlap between required skills and engineer's skill profile
Availability25%Current utilisation — engineers near capacity are scored lower
Historical performance20%Past completion rate and on-time delivery on similar tasks
Seniority fit10%Task complexity score vs engineer seniority level
Team balance5%Distribution of work across the team to avoid single points of failure

Assignment modes

Manual assignment (manager-initiated)

A manager opens a pending_assignment task, clicks Assign, and sees the AI-ranked list. Each candidate shows a confidence score (0–100%), matched skills, current workload, and a one-sentence explanation of why they were ranked there. The manager selects an engineer and confirms.

Auto-assignment

Enable auto-assignment at the project or organisation level under Settings → Automations. When active, tasks are automatically assigned to the top-ranked engineer as soon as they enter pending_assignment. You can restrict auto-assignment to tasks below a complexity threshold (e.g. complexity score ≤ 5) to keep human oversight on complex work.

Assignment acceptance flow

When assigned, an engineer receives a notification (in-app, email, and/or Slack). They can:

  • Accept — task moves to in_progress and appears on their dashboard
  • Decline — task returns to pending_assignment and the next ranked engineer is automatically suggested

Overriding the AI

Managers can always override the AI's suggestion and manually assign to any active engineer — even one who wasn't in the top suggestions. The override is recorded in the audit log, including both the AI's top recommendation and the manual choice. This data feeds back into the algorithm's historical performance tracking.

💡
Continuous learning
Every accepted assignment, completed task, and manager override contributes to the algorithm's historical performance data. Over time, Munera learns which engineers consistently deliver well on which task types, improving assignment quality automatically.

Conflict detection

Before confirming an assignment, the engine checks for potential conflicts:

  • Engineer's utilisation would exceed 110% — requires manager confirmation to proceed
  • Task has an unsatisfied dependency that blocks progress
  • Engineer is flagged as on leave or unavailable in their profile

Bulk assignment

From the Tasks list view, select multiple tasks and use Bulk Actions → AI Assign to have Munera queue all selected tasks for AI assignment simultaneously. Each task is processed independently, respecting real-time utilisation that updates as each assignment is made.