Saving Sellers over 32,000 Hours of Manual Work Using Amazon Bedrock with US Foods

US Foods is a leading American food company that serves 300,000 restaurants nationwide. The company’s sales teams work closely with customers to curate their menus and maintain their restaurant equipment. Customizing pitches for prospective customers was time intensive. Previously, sales representatives spent up to 4 hours per proposal, manually reviewing diverse data sources including paper menus, spreadsheets, and handwritten notes—some in multiple languages.

US Foods saw an opportunity to use generative AI for building a sales and prospecting application to maximize efficiency and give back time to its sellers. In 6 weeks, the company completed a proof of concept using Amazon Bedrock—a fully managed service that offers a choice of high-performing foundation models from leading AI companies. “When Amazon Bedrock came out, that just seemed brilliant to me,” says David Falck, vice president for machine learning and engineering at US Foods. “We literally played with half a dozen different models.”

US Foods collaborated with its sellers to improve the application, forming a small pilot group. “We tested with our Lake Michigan area first,” says Falck. “Then, we kept expanding the pilot and saw how other teams and other locations were using the tool. The really cool thing was that we have sellers that have been at US Foods for about 30 years and they quickly adapted to this tool.”

Considering sellers’ needs and feedback, US Foods developed the Automated Order Guide application. For relevant food and service recommendations, the company uses Anthropic’s Claude in Amazon Bedrock. The company also implemented Amazon Textract, a machine learning service that automatically extracts text, handwriting, layout elements, and data from scanned documents.

After a successful pilot phase, US Foods moved the application into production, putting it in the hands of 3,300 sellers, who immediately saw an impact. Response times per proposal fell from 3–4 hours to 20 minutes. Sellers also saved over 32,000 hours of manual work within 6 months of deployment. “They can spend more time doing valuable work such as building stronger relationships with their customers and converting more leads into customers,” says Falck.

US Foods has increased its return on investment by 10 times since deploying the application. To further streamline its sellers’ workflows, the company will embed the application into its sales system. “It’s been a fantastic journey,” says Falck. “We’re all excited about the implementation of this tool, and how it could be applied to other use cases within the company.”