Setting up WesTrac to take Generative AI ambition to reality in 6 weeks
WesTrac is a Caterpillar equipment provider with over 3,000 employees servicing the mining and construction industries. As part of their Customer Experience strategic vision, they saw the opportunity for transformation in their business using the power of Generative AI through Large Language Models (LLMs). They were looking for a tangible pathway for their Data & Analytics (D&A) team to get started. Therefore, WesTrac engaged eliiza to start the journey.
As a starting point, WesTrac set three challenges:
- Establish a Generative AI solution in production to rapidly prove value
- Validate and prioritise the use cases identified elsewhere in the business to uplift the customer experience
- Enable the D&A team with the skills, technology blueprints and processes to deploy future LLM applications
The primary candidate for production was a use case to improve WesTrac’s customer experience by automating the rewriting of invoice notes from the highly technical, shorthand language used by technicians to a more customer friendly invoice description. Rewriting customer invoices was currently being done manually by service administrators, and taking 100s of hours of manual effort per month.
To achieve these outcomes in a short six weeks, eliiza tackled objectives in parallel, working closely with the D&A team to ensure upskilling throughout. Our team used the main use case as the vehicle to set up much of the infrastructure needed for future LLM deployments. The team set up an end-to-end pipeline within the WesTrac Azure Databricks environment using the Azure OpenAI service, concurrently acting as a blueprint for further LLM automation.
To set objectives, eliiza worked closely with both the D&A team and business SMEs responsible for the invoice notes to ensure detailed requirements were documented, test cases and success criteria were established early, and business processes were clearly defined.
In parallel, the team investigated the business need and viability of alternative use cases. The one showing immediate promise was to improve safety outcomes by providing a natural language interface to query and search Safe Work Procedures (SWPs). Since Eliiza had already developed an LLM solution accelerator internally, the team were able to prove and demo the use case within two weeks of getting the initial data, in a showcase to executive leaders.
WesTrac started the 6 week engagement with an ambition to drive cutting edge LLM usage. By the end of the process, they had all of the foundations to achieve:
- Proven value by delivering LLM-generated invoice notes to service administrators via a Power BI report, creating significant efficiency gains whilst keeping a human-in-the-loop process for quality assurance
- The WesTrac D&A team were set up to undertake future LLM implementations, including the safety use case which had by then garnered broad stakeholder support
- The client had a comprehensive deployment framework to refer to for LLM usage at WesTrac including reference architectures, best practice engineering, use case validation, and ethics and risk assessment templates