Assembled transforms support operations with Claude

Assembled relies on Claude for Assist, their omnichannel AI agent and copilot. Their holistic support operations platform enables industry leaders like Brooks, Warner Brothers, Robinhood, and Stripe to deliver great support at scale, coordinating humans in partnership with AI agents.

With Claude, Assembled Assist:

• Increased customer satisfaction by 20% while decreasing support spend
• Achieved over 50% reduction in escalations
• Improved cases solved per hour by more than 30%
• Automated over 50% of cases while maintaining 90%+ customer satisfaction
• Switched to Claude in 20 minutes during a critical outage

Meeting modern support challenges with AI

Modern support teams face a challenge: how to deliver personalized, high-quality service at massive scale. The founders of Assembled encountered this challenge firsthand while building support systems at Stripe. "In the early days of Stripe, Patrick & John [Collison] would have people over to their apartment to do support tickets on rotation," said John Wang, co-founder at Assembled. "There's this perception that support is a cost center—yet every great company starts with this type of customer obsession. We've found over the years that it's a uniquely difficult problem to scale great support, as a business succeeds and gets more complex."

Assembled recognized that truly effective support requires thoughtful coordination between AI and human agents, while many competitors focus solely on automation and deflection. Wang notes, "A lot of new players are oversimplifying the support problem. They throw AI agents in the mix and think that will solve everything. We deeply understand the inner workings of customer support. Instead of deflecting low value, low complexity work, we've focused on building systems to better resolve so-called Tier 2+ complexity issues—issues that require analysis, actions, and empathy. Some of these issues can be automated end-to-end, and others involve giving human agents AI superpowers. Whatever it takes to result in a great customer experience.” Assembled’s deep understanding of support operations led them to build a product that goes beyond simple automation. Instead of replacing human agents, Assembled Assist coordinates AI and human efforts to deliver better customer experiences while reducing costs and complexity.

Why Assembled chose Claude

Assembled selected Claude for its superior reasoning, reliability, and natural conversational tone. These qualities were tested during a critical competitor outage, when Assembled needed to quickly switch their platform to maintain service for their enterprise customers. "We moved all LLM workflows over to Claude in real time, which was huge for us because we could ride out a five to six hour outage without disrupting our customers," said Wang. "Claude performed so well that we're reevaluating our entire model infrastructure. The reasoning capabilities are significantly better, and the conversational tone feels much more natural even out of the box."

Assembled discovered Claude's value for internal operations beyond customer support applications. Wang said, "When we were using Claude internally, we used it extensively to help train our sales team on our product. Early on, our sales team was using the Claude Slack bot to answer product questions during sales calls. This was so effective that we now offer all employees a monthly LLM budget—they can choose any AI model, but many prefer Claude for its natural, empathetic interactions."

The team leveraged a sophisticated evaluation system to measure AI model performance. Every support agent response goes through the evaluation system, creating a golden dataset for comparison. When agents modify AI-generated responses, the system analyzes these changes to continuously improve accuracy. This approach helped achieve dramatic improvements—for one major customer, accuracy jumped from 62% to 90% across hundreds of categories.

How Claude powers intelligent support operations

Assembled Assist uses Claude to enhance support through:

• Dynamic ticket categorization using meta-prompting that adapts to each customer's unique taxonomy
• Intelligent routing between AI and human agents based on case complexity and customer sentiment
• Multi-model architecture enabling seamless failover during outages
• Real-time response generation and enhancement
• Secure failover architecture for enterprise reliability

Transforming support for leading brands

Assembled's impact goes beyond automation metrics—they're fundamentally changing how support teams operate. Wang said, "Even the leading AI companies still use human agents for their own support. The key insight isn't replacing humans, but transforming how they work. We're focused on identifying which tasks AI should handle, how to escalate effectively, and where human expertise adds the most value."

Their approach has gained traction with industry leaders. For these companies, maintaining high-quality support while scaling globally isn't just about efficiency—it's crucial for brand reputation and customer loyalty.

• A leading consumer software company: In just the first two weeks of using Assembled Assist, they saw material improvement in both average handle time per case and CSAT. They plan on automating at least 25% of all support tickets as part of a company-wide growth push.
• Thrasio: Automates over 50% of support tickets with Assembled Assist, which led to nearly $2m in cost savings and a 50% decrease in full resolution time.
• Honeylove: Saw massive productivity and efficiency benefits after just 5 months using Assembled Assist – solves per hour increased by 54%, escalations decreased by 20%, and CSAT for AI-powered responses increased by 18%.

Building the future of AI-enhanced support

Assembled is developing features to help companies deploy the right resource at the right time, using Claude 3.5 Sonnet for complex reasoning tasks and Claude 3 Haiku for high-speed operations. Their platform analyzes the skills required for different types of issues and automatically routes cases to the optimal handler—whether that's an AI agent or a human with specific expertise.

Wang concludes, "We're not just playing a deflection game, because support is more complicated than that. It’s also time-consuming and expensive work, especially at a global scale. We're building something that actually powers better, faster work for support teams and their customers; for example, we’re working on leveraging LLMs to take language barriers out of the global support delivery. Through their work with Anthropic, Assembled aims to make this vision a reality—transforming customer support from a cost center into a strategic advantage that builds stronger relationships between companies and their customers.