Why data readiness matters more than the chatbot
A leading Aussie enterprise had a growing library of reports and metrics across multiple business functions, making it difficult for staff to quickly find the right information.
Skillfield embedded consultants within the reporting and insights team to build the data foundations for a natural-language AI assistant, creating a structured report catalogue, data dictionary and metadata framework that now enables intuitive, self-service access to the organisation’s reporting landscape.
The Problem
Like many large organisations, the reporting and insights team had, over time, built and maintained a broad and growing library of reports spanning multiple business functions.
However, there was no single reliable way for staff to understand which reports existed, who owned them, where they were located or which metrics and datasets they used. Access also varied between reports, adding another layer of complexity.
Finding the right metric or report often depended on the reporting team’s knowledge of the environment. This created an unnecessary bottleneck and placed additional demand on the reporting team.
The Solution
Skillfield consultants helped design and develop a natural-language chatbot (AI assistant) that lets staff ask plain-language questions, such as which reports contain a particular metric, and receive a direct answer.
Making this possible required more than building the assistant itself. Before the assistant could work reliably, the underlying reporting landscape had to be made machine-readable. This meant cataloguing existing reports by their owner, business function, and location, building a data dictionary that defined what each data field represents, and tagging the relationships between tables and data sources so the assistant could correctly interpret everyday business language. The AI assistant responses were then refined through iterative prompt engineering to improve the accuracy, relevance and reliability of the user experience.
The Outcome
The AI assistant is now fully functional and used by business teams across the organisation, giving staff a simpler way to navigate the reporting landscape. Rather than relying on the reporting team to locate information, users can ask the assistant directly.
The value also extends beyond the assistant itself. The report catalogue, data dictionary, and structured metadata created for the solution now provide a reusable data foundation, supporting stronger information discovery today while positioning the organisation to explore future reporting, analytics and AI use cases more efficiently.
