Successor Planning
- Year
- June 22, 2026
- Role
- AI Engineer
- Client
- PT. Mitra Pinasthika Mulia
- Access
- Internal only
- Python
- AHP
- TOPSIS
- LangChain
- vLLM
Internal system — no public link or screenshots.
What follows describes the problem, the approach and the engineering decisions at a level that discloses no proprietary detail.
A decision-support tool for HR succession planning. Given a pool of candidates and a set of weighted criteria, it produces a defensible ranking of who should be considered as successor for a role.
What it does
- Structured ranking: pairwise criteria weighting through AHP (Analytic Hierarchy Process), then candidate ranking through TOPSIS.
- Consistency checking: AHP's consistency ratio flags when a decision-maker's pairwise judgements contradict one another, before those judgements reach a result.
- Scenario simulation: an LLM layer explores variations, such as how the ranking shifts if a criterion is reweighted or a candidate's development gap were closed.
Why two methods rather than one model
The classical methods do the deciding and the LLM does the explaining. AHP and TOPSIS are deterministic: the same inputs always produce the same ranking, and the arithmetic can be shown to a reviewer. That is a hard requirement for a decision affecting someone's career, since a ranking nobody can reproduce or audit is not usable in an HR process.
The LLM never touches the score. It is used only to narrate scenarios around a ranking that has already been computed.