The global online travel company cut supplier analysis from a full quarter to days while uncovering data gaps that improved the accuracy of its emissions footprint using Watershed agents.
Booking Holdings' Lauren Esposito used Watershed agents to automate the company's supplier engagement analysis—cutting a full fiscal quarter of manual work to days and uncovering data gaps that shifted a brand's footprint calculation by several percent.
Challenge
Analyzing supplier data for scope 3 engagement consumed a full fiscal quarter across multiple team members and consultants, relied on stale spend data, and left critical gaps undetected.
Solution
Esposito built an AI-powered workflow using Watershed agents that reconciles supplier records, compares spend and emissions signals across years, and recommends engagement tiers with rationale.
Results
What took three people plus consultants a full quarter now takes days. Esposito also identified 7,000 records missing regional labels, corrected them, and adjusted one brand's footprint calculation by several percent.
Challenge
Booking Holdings—the company behind Booking.com, Priceline, KAYAK, and OpenTable—operates across more than 220 countries, with different global headquarters for each brand and a multi-brand emissions footprint to match. The company lists 4.7 million properties and employs 24,000 people across more than 240 offices in 100 countries—and its data reflects that complexity, arriving in different formats, currencies, and languages. “From a sustainability data perspective, we’re bringing together information from different brands, systems, suppliers, and regions,” explains Lauren Esposito, the company's Sustainability Data & Analytics Manager.
Esposito is responsible for the data behind Booking Holdings' Vendor Engagement Program—the analysis that determines which suppliers get prioritized for scope 3 engagement and target-setting. Every cycle, that analysis would consume a full fiscal quarter for Esposito, two program managers, and external consultants. “Our data was spread across the platform and it was hard to build a workflow that provided scale and consistency year over year,” she explains. “We need to get this dataset right, so we prioritize the most impactful suppliers for engagement.”
The process had other problems, too. Vendor segmentation was updated just once a year from stale spend data. Grouping adjustments took weeks, supplier records sat in disconnected datasets with no reliable way to reconcile them, and critical gaps like missing regional labels went undetected, silently skewing footprint calculations. Before agents, just navigating the data meant opening twenty browser tabs, each with a different footprint drilldown.
Esposito wanted to fix this. When she joined the inaugural cohort of Watershed's AI Fellowship in summer 2026, she came with a clear target: a to-do list of things she had no idea how to do—and the intention to conquer each one.
“Since getting access to the Watershed agents, I’ve basically taken my to-do list of things I had no idea how to do and tried to conquer them one by one. I started with the things that had been sitting on the list for months because they were either very manual or difficult to know where to begin. It’s been really fun.”
Lauren Esposito,
Sustainability Data & Analytics, Booking Holdings
Solution
Using Watershed agents, Esposito built a workflow that pulls procurement data from multiple Watershed datasets and reconciles supplier records into a single coherent view—something her team had never been able to do reliably. “I used to open twenty tabs, each with a different footprint drilldown,” she explains. “Now I have one dataset combining methodologies, year-over-year trends, and disclosure data—all pulled together from different modules in the platform."
For each supplier, her AI agent compares spend and emissions across years and methodologies, checks SBTi and disclosure status, and flags anomalies, like high-emitting vendors whose emissions rise while spend stays flat. It then recommends tranche placement with rationale, proposing segmentation based on real signals rather than static rules.
What used to require weeks of manual cross-referencing across disconnected spreadsheets now happens in a few minutes. “The agent is helping with the first pass of organizing the information,” explains Esposito. “My team can then focus on reviewing the output, deciding whether the segmentation makes sense, and determining what kind of engagement is appropriate for each group.”
Results
Esposito compressed a full fiscal quarter of work—across three team members plus consultants—into days.
But the speed was only part of it. With her new workflow, Esposito identified 7,000 records missing regional labels and mapped them to the correct global headquarters. "I'd had it on my action items for months—figuring out what country each of these suppliers came from,” she says. “I had no idea how to do it. With the agents, I did it in minutes."
“It’s not just about speed, it’s about making things that weren’t possible, possible. That is the real goal: the agents help us get to better information faster, so we can focus our supplier engagement where it can improve the quality of the footprint.”
Lauren Esposito,
Sustainability Data & Analytics, Booking Holdings
Where the company had previously defaulted to US estimates for suppliers it couldn't cleanly geo-locate, she can now separate Singapore from Indonesia and apply accurate regional factors immediately. That accurate regional mapping revealed the company had been undercounting: she adjusted one brand's footprint calculation upward by several percent.
The workflow also changed how often the analysis can happen. Vendor segmentation used to be updated once a year, whereas Esposito is now aiming for four times a year.
And the impact extends beyond her own desk. "The agents are for everybody," Esposito says. "I'm really looking forward to this scaling out to the rest of my team, because everyone will be on it."