Valrhona processes more than 100,000 customer orders a year from pastry chefs and distributors across over 120 countries. Until recently, most of those were keyed into SAP by hand.
That meant customer service staff opening emails, reading attached PDFs, and manually entering every line into SAP. About two and a half minutes per order. Multiply that across more than a hundred thousand orders and you get a team that spent most of its time on data entry instead of actually talking to customers.
Valrhona deployed Esker's AI-powered order management tool, which reads emailed purchase orders, extracts the relevant data, and feeds it into SAP automatically. 95% of emailed customer orders now process without anyone touching them. Average time per order fell from roughly two and a half minutes to 46 seconds, with significantly fewer data entry errors.
That 95% applies to orders received via email, not every order across every channel. But email was a major source of order volume, and clearing that bottleneck changed what the team actually spent their days doing.
"Esker has allowed us to shift our approach," Perrine Guironnet, Head of Customer Services at Valrhona, said. "We now use the time we've saved to anticipate needs, follow up with and better track our customers' activity."
The customer service team went from entering orders to managing relationships — anticipating needs, following up with customers, and better tracking their activity across Valrhona's professional client base.
The technology here is not glamorous. There's no frontier model, no autonomous agent orchestrating sub-agents. Esker captures order data from documents and integrates it into ERP fields. It's the kind of AI deployment that will never make a keynote but it saved a century-old company thousands of hours across a global operation.
The most effective AI deployments happening right now are ones nobody wants to talk about at conferences. Reading emailed purchase orders, matching SKUs, filling in database fields. Valrhona freed up its customer service team with technology that's been quietly available for years, and the only reason it makes news is that most companies still haven't done the same. If your team is still manually entering data from one format into another, the frontier model isn't what you're waiting for.
