Leveraging GenAI to classify and unlock a helicopter view of a hotel chain’s expenses

For this project, we worked with a Hotel chain that was aiming to optimize its expenses which were distributed across several locations and multiple stakeholders. The client struggled to centralize information as the hundreds of thousands of expense items were not standardized and were hard to analyze.

This is oftentimes a real challenge that our customers face when starting data projects, where information is unstructured, which blocks further steps of analysis and a true insight-driven decision-making process of optimization.

Additionally, the purchasing process was scattered across the organization where each person involved conducted their own practices within a certain range of freedom, which also created complexity in finding real and tangible actions to take to optimize these processes and the underlying expenses.

Finally, the filled fields of information were frequently inconsistent among each other (e.g.: numerical quantities were filled as individual units, other times as kilograms, liters or packs), which created doubt in existing analysis within the teams, hindering a clear action plan to tackle these goals.

As with most of the projects we conduct at LTPlabs, we aim to take a holistic approach to data projects. Our experience tells us that problems with data are born in other places such as tools, people, processes, etc.

In the face of the uncertainty the team felt regarding the existing data sets, we needed to make sure that there was a common understanding of metrics, how they were captured and identify opportunities for improvement. This created a baseline scenario where we could be comfortable with the numbers and could standardize the data sets to make them comparable.

‍

This full process mapping involved stakeholders from across the organization, creating a common language between them and tapping into existing knowledge regarding the data.

Having the processes mapped and a good grasp of what the existing data meant, we could now start working with the data. In order to clean up and standardize the data, we leveraged Generative AI models in a RAG system that accessed financial and accounting standards for the hotels sector, which gave the models a higher precision and speed.

Generative AI solutions allow us to perform analysis that were just too costly to perform before. Labeling hundreds of thousands of expenses by hand would be unthinkable as we did not yet know how much value was in the data we were looking at, yet now, with an off-the-shelf solution and some fine tuning, we can do it in a matter of hours.

This layer of analysis, unlocked by GenAI, is the “boring” side of it that nobody talks about, yet often unlocks the most value, allowing companies to take advantage of Analytical AI is new ways – in this case, to optimize millions of euros in expenses.

The project’s goal was to quantify the potential savings in expenses across the different hotel locations. In order to enable it, we divided the analysis in more detailed rubrics where there could be a potential change in action (eg.: negotiating contracts, streamlining expenses in contracts, replacing ingredients for lower cost alternatives, etc).

This analysis tapped into descriptive analytics methodologies, embedding existing business metrics and industry benchmarks as a complement, creating scenarios with different confidence levels.

When looking at the potential of something, we like taking a conservative and realistic approach, which makes the project team comfortable with the numbers they obtain. It give everyone a baseline approach which is based on real numbers, can be acted upon and, most importantly, people believe in them.

Concluding, the analysis allowed us to develop a comprehensive report which served as a blueprint for expense optimization in the chain, including:

• A complete process mapping from central to local teams, flows of information, tools used and points of validation with flagged pitfalls and suggested improvements.
• A thoroughly AI cleaned and labeled data set of expenses that spanned the past 2 years activity in all hotel locations,allowing for future use.
• A detailed analysis of this data, divided by action taken and with corresponding levels of confidence, allowing the companyto add or remove rubrics as needed.
• A recommendation of 14 actionable cost-saving measures that the company can take immediately, shortening the path to ROI from this project.

‍“We’ve been discussing the potential of savings for years but only had our intuition to guide us. Now we’re able to know exactly how much potential there is with an accurate and scientific basis!” - Client

‍

Through this project, the company was able to identify a potential reduction of 5% of its expenses across all hotel locations with clear initiatives and a roadmap of AI initiatives that can support further expense reduction with operational tools that enable this transformation.