The problem: rostering is genuinely hard
Building a ward roster means satisfying simultaneously hard constraints (rest periods, minimum staffing, skills) and preferences (fairness, requests, sustainable sequences). This is the Nurse Rostering Problem, studied for decades by operations research precisely because the combinations grow explosively: adding one person or one rule multiplies the cases to evaluate.
The idea: the “annealing” of metals
Simulated Annealing takes its name from the heat treatment of metals: the material is heated and then cooled slowly so that its structure settles into the best form. The algorithm does the same with the shift plan: it starts from a solution, modifies it by trial and — while the “temperature” is high — it also accepts temporary setbacks. That is its decisive advantage: it does not stop at the first passable solution, but keeps exploring until it finds a genuinely good one. As the temperature is gradually lowered, the plan stabilizes on the best configuration found.
Why “born from research” is not a slogan
Windex promoted and funded a research project with the University of Udine to adapt the method to real-world rosters in Italian healthcare: national collective agreement (CCNL) constraints, on-call shifts (guardie), part-time contracts, public holidays. The result was published in 2023 in a peer-reviewed scientific journal, in open access: anyone can read how the engine works — including your CIO — and its performance has been put to the test on the benchmarks of the international INRC competitions, the proving ground of the scientific community.
And artificial intelligence?
The ALISEO ES engine is artificial intelligence — of the right family for this problem: optimization, not text generation. The difference that matters to Management is not the label, but verifiability: every plan produced can be checked constraint by constraint, the method is published and the results are reproducible. A documented, verifiable AI, working in the service of the people who then sign off on the roster.
→ Read more: Shift optimization
Reference: Ceschia, Di Gaspero, Mazzaracchio, Policante, Schaerf — “Solving a real-world nurse rostering problem by simulated annealing”, Operations Research for Health Care 36 (2023) 100379, open access — doi.org/10.1016/j.orhc.2023.100379