AI & Machine Learning Policy
Understanding how our rostering system uses AI
Overview
SKEDA uses agentic AI, providing it with optimisation algorithms tooling to solve complex scheduling problems. AI and LLMs are used only as a coordination function, automatically evaluating the quality of the solutions before they are shown to our users, in order to provide the best possible experience. All the raw data generated by the solving algoritms is shown to the user, so that the agent cannot inadvertedly limit their control over the solutions proposed.
Our Technology
Tabu Search
We use a sophisticated optimization algorithm called Tabu Search. This technique is very effective at solving complex scheduling problems where multiple constraints must be satisfied simultaneously.
Why Not Large Language Models (LLMs)?
Inefficiency
LLMs are not designed for mathematical optimization and would be computationally expensive for scheduling problems.
Lack of Precision
LLMs generate text, not mathematical solutions, making them unsuitable for precise scheduling optimization.
Constraint Handling
LLMs cannot systematically handle complex constraint systems required for legal compliance.
Our Approach
Purpose-Built
Our algorithm is specifically designed for scheduling optimization problems.
Computationally Efficient
Optimized for speed and resource usage, making it practical for real-world applications.
Constraint-Aware
Systematically handles complex compliance requirements and staff preferences.
Important Limitations
No Guarantee of Optimality
While our algorithm finds good solutions, it does not guarantee finding the absolute best schedule. This is a fundamental limitation of heuristic optimization approaches. However, our solutions are consistently high-quality and meet all specified constraints.
Transparency & Control
We believe in transparency about our AI systems. You can review the constraints and preferences that influence our scheduling decisions, and our system provides explanations for why specific assignments were made. This helps build trust and allows for human oversight and adjustment when necessary.
Questions?
If you have questions about our AI and machine learning approach, please don't hesitate to contact us.