Enterprise Knowledge Search · Sales Automation
Mattress Firm wanted to improve frontline productivity, sales conversations, and job satisfaction for Sleep Experts, but first had to define the right features, prove the business case, and build the technical infrastructure for a reliable generative AI assistant.
Reported impact
Mattress Firm reduced the time to launch an MVP, gave frontline teams faster access to answers and training, lowered reliance on help-desk calls, and scaled the assistant from a pilot group to thousands of employees across its US store network.
Mattress Firm worked with Bain and Google Cloud to build SleepExpert.AI as a production assistant for frontline retail staff. In 12 weeks the team launched a conversational system that combines chat and voice access with a large knowledge base and realistic role-play training. Drill Mode uses more than 2,000 synthetic personas to simulate customer conversations, while the agentic AI Drill Builder lets the retail training team create and edit scenarios quickly instead of managing them manually. Chat Mode answers questions about products, promotions, store operations, and technology issues so Sleep Experts do not need to rely on help-desk calls or memory alone. The deployment also added governance mechanics such as beta-user testing, a product backlog, a Knowledge Council, and KPI tracking. The source does not spell out a storage topology, but it does make clear that Google Cloud provides the generative AI and cloud-storage foundation for the assistant.
Product, promotion, operations, and troubleshooting knowledge flows into SleepExpert.AI on Google Cloud. Sleep Experts interact through chat or voice, Chat Mode returns answers from the knowledge base, and Drill Mode plus the agentic AI Drill Builder generate synthetic training sessions that the retail learning team can update continuously.
"Chat Mode provides real-time, relevant answers to critical product information questions"
"In just 12 weeks, the team built and deployed SleepExpert.AI"
"The training simulates critical customer interactions using role-play with over 2,000 synthetic personas"
"fewer calls to a help desk, faster upskilling of newer Sleep Experts"
"rolled out to over 6,000 employees across 2,300 US stores"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
LangChain Sample Feb 14, 2024 Rakuten Group Rakuten wanted to augment both business clients and employees with AI products for market analysis, support, and knowledge work while keeping control over prompt quality, model experimentation, and enterprise data… improve productivity by 20% LangChain · LangSmith Hub · LangSmith Testing and Eval 
Bain Sample Archived · Oct 2, 2026 Magalu Magalu wanted to make product discovery, comparison, purchase, and post-sale support faster and more intuitive, replacing complex retail navigation flows with a conversational shopping experience that could still… 300,000 sellers Lu · Lu's WhatsApp · WhatsApp 
Google Cloud Sample Archived · Sep 17, 2026 VELLOWS VELLOWS needed a secure AI platform that could standardize knowledge management and accelerate proposal writing, but strict privacy rules and disconnected expertise repositories kept consultants from safely using AI at… Achieves 14x faster proposal creation Agent Designer · Agent Search · Cloud Run 