Product descriptions:
Lippert is a $3.8 billion global manufacturer that supplies critical components for RV, marine and automotive brands. As demand for mobile living surged, so did the pressure on Lippert’s data, support and analytics operations — especially when call volumes spiked into the millions and legacy tools like Synapse and Dynamics 365 couldn’t scale. Fragmented systems, slow onboarding and limited insight into service quality left the team reactive and resource-constrained. With Databricks and Mosaic AI, Lippert consolidated their data into a single lakehouse and began deploying intelligent agents that help customer service teams resolve issues faster, coach agents more effectively and automate insights across the enterprise. What once took months of manual effort is now automated, measurable and repeatable — paving the way for future agents in HR, warranty and supply chain.
Legacy data warehouses put the brakes on customer care transformation
As a global manufacturer of components for premium RV, automotive and marine vehicles, Lippert plays a foundational role in outdoor recreation and transportation. With over 12,000 employees and 140 facilities worldwide, the company powers more than just products — they power experiences. “We supply more than 50% of the content that goes into an RV,” Kenan Colson, VP of Data and Artificial Intelligence at Lippert, explained. “But we don’t build the RV itself. That means if there’s an issue, the root cause could be the product or how it was installed. We need data and insight to tell the difference.”
To meet rising customer expectations and modernize operations, Lippert invested in a series of AI-powered workflows, starting in the area that needed it most: customer support. That pressure first peaked during the pandemic, when RV sales skyrocketed and Lippert’s call center was overwhelmed by a new wave of customer inquiries, logging over a million customer touches per year. At the time, onboarding a new support agent could take six months, given the complexity of Lippert’s product lines. The team needed a way to scale knowledge and service quality fast.
That’s why they began building an AI assistant, trained on product manuals, technical case history and YouTube videos from Lippert’s field experts. This customer support center chatbot now supports agents in real time, surfacing relevant troubleshooting guidance based on the customer’s situation and product. According to Kenan, “This is going to cut that six-month timeframe to at least half,” referring to how much faster new agents can get ramped up with the AI assistant by their side.
Lippert also turned to AI for service quality and coaching. Previously, they relied on a third-party firm to manually review and score just 100 calls per month, looking for key sales behaviors and process adherence. That limited visibility made it difficult to coach agents consistently. Today, Lippert uses AI to analyze thousands of calls each day, scoring performance across agents and product lines and identifying who may need retraining. “Now, we’re doing all the calls, not just a small random sample,” Chris Nishnick, Director of Artificial Intelligence at Lippert, said. “We can help the training team identify exactly who needs support, whether it’s trouble resolving issues or struggling to offer upsells.”
Lastly, the team deployed automated call summarization and transcription, giving every agent quick access to the full history of a customer’s interactions. Summaries are pushed directly into Salesforce CRM, allowing any agent to step into a conversation and pick up right where the last one left off. Kenan explained the value of this, saying, “Instead of having to read through 10 previous call summaries, we’re moving toward case summaries that give agents exactly what they need — right as they start a call.”
These innovations weren’t possible on Lippert’s legacy architecture. Their previous data warehouse, Synapse, quickly became a bottleneck. “It was getting cost-prohibitive. We had already maxed it out, and the next tier was going to cost us another $18,000/month. And we were only 40% of the way there,” Kenan said. “It just wasn’t sustainable to adopt these new use cases.” Beyond cost, Lippert dealt with technical limitations, siloed systems and a lack of scalability. With fragmented data spread across on-premises systems and Microsoft Dynamics, it was hard to unify the data to deliver actionable insights based on comprehensive analytics. The fragmentation also made it difficult to run concurrent queries at scale, further limiting the team’s ability to deliver real-time insights. “Before Databricks, doing GenAI at scale would have been way too manual and resource-intensive,” Chris said. The team couldn’t move fast enough to support the business. That all changed with their move to the Databricks Data Intelligence Platform.
Building a customer support agent with Databricks Mosaic AI
To build their customer support AI assistant and automate call and case summarization, Lippert turned to Databricks Mosaic AI’s unified framework to streamline every step of the AI lifecycle.
One of the earliest breakthroughs came when Lippert began developing their customer support AI agent using retrieval augmented generation (RAG). Building the prototype wasn’t the hard part — validating that the agent could reliably answer customer questions was. “That’s where things would stall out. We had anecdotal evidence, but no measurable way to confirm it was answering questions correctly,” Chris recalled.
That changed with Mosaic AI. Using their evaluation framework, Lippert generated a synthetic dataset of 50 realistic question-answer pairs based on prior case history. This allowed the team to run performance tests and resolve gaps in retrieval and chunking, before ever involving subject matter experts (SMEs). “We improved model accuracy from 33% to 56%, just using synthetic data,” Chris said. “We brought in the SMEs around 54% to help us fine-tune from there.”
This hybrid approach accelerated training and helped build SME trust in the model’s performance. Once deployed, Lippert used Mosaic AI’s production monitoring tools to track every live request. “We can measure and monitor accuracy, latency, safety and groundedness in real time,” Chris added. Unity Catalog ensures secure, role-based access to sensitive information, giving Lippert confidence in their data governance. Meanwhile, MLflow manages the full lifecycle of AI experimentation — tracking pipeline versions, evaluation scores and model changes so teams can collaborate efficiently and iterate with confidence.
For business users, Lippert has begun experimenting with AI/BI Genie. Instead of relying solely on static dashboards, internal teams can ask natural language questions of the data, turning self-service analytics into a conversation. Teams can explore everything from customer sales trends and sentiment to recurring support issues, all without writing a single line of SQL. As Kenan explained, “We haven’t productionalized it yet, but we really like using AI/BI Genie. It’s great to be able to ask questions about the million calls that are out there. It’s a lot of unstructured data that’s not very easy to query otherwise, but Genie is doing a really good job.”
Lippert has also leaned into Databricks’ broader data and productivity ecosystem. By unifying all their data into a lakehouse, all teams — technical and nontechnical alike — now operate from a single source of truth. With data centralized and accessible, the company can shift their focus from infrastructure management to innovation. Analysts are empowered to move faster, business users get timely insights and AI teams can build with confidence, knowing the data foundation is reliable.
Delta Lake allows the team to unify data across multiple warehouses, while Databricks SQL supports thousands of reports visualized through Power BI. Ephemeral cluster management — an approach where compute clusters are spun up temporarily to handle specific workloads, then automatically shut down once the job is complete — gives them elastic compute that scales on demand. As more teams get involved in AI, tools like Databricks Assistant are helping boost productivity across the board. Assistant lets users query data through a conversational interface. Kenan herself leverages it to write code, recently asking it to parse a timestamp from a file name using regex.
With Mosaic AI at the center, Lippert has transformed their ability to build and deploy GenAI agents at scale. Even better, they’ve created a repeatable, cross-functional development model that empowers engineers and SMEs to work together in a shared feedback loop. According to Chris, “We’re on the same journey now. They’re invested. We’re invested. And we’re building something better, together.”
To help accelerate this momentum, Lippert is using Mosaic AI Agent Bricks — a tool specifically designed to build and optimize agents on your data and your task. It streamlines everything from synthetic data generation to evaluation and grounding loop feedback, helping teams go from prototype to production faster.
“With Agent Bricks, we can quickly productionize domain-specific AI agents for tasks like extracting insights from customer support calls — something that used to take weeks of manual review,” Chris said. “It’s accelerated our AI capabilities across the enterprise, guiding us through quality improvements in the grounding loop and identifying lower-cost options that perform just as well.”
AI agents that pay for themselves: $2.1M in projected savings, 106K hours saved
With Databricks and Mosaic AI, Lippert has accelerated their GenAI delivery while achieving measurable returns across customer support, analytics and workforce efficiency.
The cost savings from consolidating their fragmented data landscape into a single platform have been substantial. By integrating with critical systems like Microsoft Dynamics 365 and sunsetting legacy systems like Azure Synapse and on-premises SQL Server, Lippert expects to save $3.83M per year.
The AI-powered customer support chatbot, now live across multiple product lines, is “killing it,” according to Kenan. “It’s expected to augment 50% of all inbound call volume. This is a major unlock for a call center that receives more than 1 million calls per year. By the end of 2025, the company expects to save $2.1 million, reclaim 106,680 hours and generate capacity equivalent to 51 full-time employees — all without expanding headcount.
At the model level, Lippert has seen dramatic performance gains. Starting at just 33% model accuracy, the team used Mosaic AI’s synthetic data and evaluation loop to reach 84% accuracy in just weeks. The AI assistant has also revolutionized internal onboarding, reducing ramp time for new agents from six months to just four weeks — an 85% reduction. These improvements in knowledge delivery and training have directly enhanced customer service performance.
Across the organization, Lippert’s data team has seen a 40% productivity lift. They expect to unlock substantial annual productivity dividends by reducing manual work and accelerating delivery timelines. With faster data access and a unified platform, Lippert now delivers key business insights in days instead of weeks, driving a 43% acceleration in time to market for data-driven initiatives.
Encouraged by these wins, the team is actively expanding their GenAI footprint across the enterprise, standing up new agents for HR, warranty operations and supply chain optimization. The HR agent is designed to streamline internal support. It will automatically route employee questions to the right inbox and handle basic requests like benefits FAQs. With multiple HR inboxes currently managed manually, Lippert anticipates this agent will significantly reduce back-and-forth and offload routine tasks. “They don’t always email the right distribution list,” Kenan said. “Having AI route requests — and even answer some of them — is a big win for both speed and efficiency.”
The warranty intelligence agent will help Lippert identify early product issues by mining customer call transcripts for patterns. Warranty analysts currently review up to 1,200 claims per week, often manually scanning descriptions to determine whether the issue was with the product or a component, and whether it was installed correctly. “Instead of waiting until a claim comes in, we want to catch it at the call center level. The earlier we can detect something, the more we can avoid future claims and recalls. That’s where the ROI really kicks in,” Chris explained.
The most financially transformative initiative may be the supply chain agent, which will focus on vendor and contract optimization. Today, Lippert’s procurement team only has bandwidth to review the top supplier’s spend, leaving a long tail of smaller suppliers unchecked. The agent will cross-reference contracts, commodity indexes and exchange rates to suggest smarter sourcing strategies. According to Kenan, “Our Chief Supply Chain Officer sees up to $12 million in potential savings. Even if we conservatively cut that in half, it’s still an 11x ROI use case.”
The vision doesn’t stop at internal transformation. Looking ahead, Lippert plans to expose their support agent externally to dealers and OEMs, giving them real-time troubleshooting guidance in the field, without needing to call the care center. With a unified data platform, a repeatable agent development process and a growing library of use cases, Lippert is building a scalable foundation for AI — one that ensures every customer interaction is faster, smarter and more personal than ever before.