AI OJT Leadership Assessment

Problem

Assessing a manager's people-development ability often stays subjective. Raters may rely on recent incidents, give high scores to every dimension because of one strong area, or rate themselves too generously. This skill turns OJT (On-the-Job Training) leadership review into a structured workflow that links scores to concrete examples and a fixed set of dimensions, reducing low-signal conclusions such as “not bad” or “needs improvement.”

How It Works

The core capability is evaluating a manager's ability to develop direct reports across five dimensions:
- Business instruction and unified leadership: whether tasks are clear, responsibility is owned, and team members are comfortable raising issues.
- Planned coaching and capability development: whether development plans, developmental tasks, one-on-one coaching, and growth reviews exist.
- Participation and goal alignment: whether team members contribute to goal setting and understand how their work supports department targets.
- Motivation and morale: whether the manager understands different motivation drivers, gives timely feedback, and rebuilds confidence after setbacks.
- Cross-department collaboration: whether the manager maintains external relationships, resolves conflicts, and secures needed support.

The workflow follows an S0-S6 state machine: confirm intent, collect the evaluated person's name, role, rater perspective, and rater name, then verify prerequisites such as at least one direct report, at least three months of observation, and a clear rater perspective. During evaluation, the skill processes one dimension at a time, asks for concrete examples, and maps them to the 1-10 rating scale. Each dimension passes a consistency gate: if the examples are strong but the score is low, or the examples are moderate but the score is high, the skill prompts the user to revisit the judgment. Before report generation, it also scans for halo effects, flat scoring patterns, self-assessment bias, and recency bias. The final report is written as .docx via scripts/generate_report.py, with Markdown/HTML fallback if the script or dependency is unavailable.

Boundaries

It is intended for mid-level managers who lead at least one direct report, especially in supervisor, peer, or self-assessment contexts. It is not suitable for individual contributors without direct reports, purely technical skill evaluation, or full 360-degree leadership reviews. It does not produce legally binding performance-review documents and does not replace professional talent assessment or psychological counseling. Results depend on observation depth and example quality, so they should be treated as development discussion material rather than the sole basis for promotion or compensation decisions.