Warranty requests at the country’s largest Christmas tree brand used to take a day to process. They now get a response in under a minute.

National Tree Company is the country’s largest importer of artificial Christmas trees, selling more than 4,500 SKUs through major retail partners and its own direct-to-consumer channel, and moving up to 60,000 orders a day at peak. Under a CEO recruited to accelerate the direct-to-consumer side of that business, customer service had become one of the more visible constraints on the brand, since that experience is most of what a customer encounters after purchase.

The operation ran manually through Outlook, with no ticketing system, no tagging, no AI, and no structured view of what customers were even asking. Through the November and December peak they worked seven days a week at time-and-a-half, carried a backlog of roughly 3,000 emails, and abandoned more than a third of inbound calls, and because the category compresses a year of demand into eight weeks, a service failure in November can’t be recovered in February. Warranty claims carried financial exposure on top of that, which raised the cost of getting an automated answer wrong.

National Tree came to us expecting a chatbot on the website, and the assessment we ran at the outset pointed somewhere else: the constraint was not the technology but the team, a small customer service group with limited bandwidth and no exposure to working alongside an AI agent. We moved the pilot to email, where customers expect a reply in days and where a weak AI response can be caught and corrected before it reaches anyone else.

We began by analyzing 66,000 call transcripts and then running Intercom’s Fin silently against the live inbox for six weeks, tagging every inbound message against a preset taxonomy, which established that warranty requests accounted for 37% of real customer volume at 94% tagging accuracy and made warranty the pilot use case. We scoped the build to the company in front of us, since the enterprise platforms in this category can cost half a million dollars a year while Fin is sized for a business National Tree’s size, so the work was workflow and prompt engineering: we rewrote the warranty policy out of legal language into plain English structured for retrieval, built a Shopify connector for order lookup, and set up human-in-the-loop for financial decisions.

Across six weeks of live production, Fin handled warranty email at 91% response accuracy and a 60 to 65% blended resolution rate, easily meeting industry benchmarks. Customers noticed the speed first, since a warranty inquiry that had previously waited a day or more for a first reply now came back complete and accurate in under a minute, including at nine at night, from a desk that staffs a few hours a day in the off-season. Spanish-language inquiries came back in Spanish and reached the team in English, negative sentiment on warranty conversations fell against the pre-pilot baseline, and several customers wrote back unprompted to thank the agent by name.

National Tree entered the pilot with no AI anywhere in customer service, and modeled against a benchmark $8 per human-handled resolution, the 40% rate we recommended targeting before peak is worth six figures a year. We’ve since extended the engagement beyond the pilot and are continuing to tune the system while their new customer experience manager is developed into the internal owner of it.