In a recent LinkedIn post, Teresa Torres discusses the intricate challenges and innovative solutions involved in developing AI customer support agents, particularly focusing on the capabilities of the startup Lorikeet. Torres highlights Lorikeet’s mission to create an AI that not only resolves customer issues efficiently but also possesses the crucial ability to recognize its limitations and gracefully hand off complex queries to human agents.
Torres emphasizes the journey Lorikeet undertook, moving past initial concepts like reflection tools and information dashboards to address the core need of businesses: clearing customer inboxes. She notes the evolution from a basic command-line script to a sophisticated system now running two distinct agents: a Concierge for direct customer ticket handling and a Coach for enabling customers to configure, test, and improve the AI’s performance.
The Crucial Role of “AI Humility” in Customer Support
A central theme in Torres’s coverage is the concept of “AI humility,” which she explains is a core design principle for Lorikeet. This principle dictates that the AI should default to seeking human assistance when it encounters situations beyond its current capabilities. This approach is vital for maintaining customer trust and ensuring effective problem resolution, especially in regulated industries where accuracy and compliance are paramount.
“Lorikeet’s vision: an agent that responds like the best customer support you’ve ever had — one that knows you, gets things fixed, and hands off gracefully when it’s out of its depth.”
Torres elaborates on how this “AI humility” is implemented, detailing the development of customer-configurable guardrails. She shares an anecdote about how a cannabis company’s unique support ticket requirements initially challenged their approach, underscoring the need for domain-specific configurations.
Human-AI Collaboration and Configuration Workflows
The article further delves into Lorikeet’s innovative “resolution in the loop” pattern, a mechanism designed for seamless human-AI collaboration. This pattern allows human agents to unblock the AI without taking over an entire ticket, optimizing efficiency and agent workload.
“The team spent months exploring the wrong ideas — reflection tools, information dashboards — before a healthcare startup pulled them toward the real problem: just help us clear the inbox.”
Torres also highlights how Lorikeet is rethinking the configuration workflow itself. Instead of customers defining processes first, Lorikeet is shifting towards a model where customers define what constitutes a successful resolution before any standard operating procedures are established. This customer-centric approach ensures the AI is aligned with business objectives from the outset.
Evolving User Experience for AI Configuration
The post touches upon the user experience challenges in configuring AI systems. Torres points out the evolution from a traditional workflow builder to a more conversational interface, yet acknowledges the persistent difficulty users face when interacting with a blank chat box, a common hurdle in AI product design.
“Why ‘AI humility’ — defaulting to human handoff when uncertain — is a core design principle.”
Furthermore, Torres notes Lorikeet’s strategy of integrating with existing platforms like Zendesk and Intercom, rather than aiming to replace them. This approach allows businesses to leverage their current tools while enhancing their customer support capabilities with AI.
A Culture of Customer-Centric Learning
Teresa Torres underscores the product engineering culture at Lorikeet, where every engineer is tasked with learning something new from a customer on a weekly basis. This commitment to continuous customer feedback is presented as a key driver of their product development and innovation.
“How customers define their own evals and guardrails through the Coach interface.”
In essence, Torres’s report provides a comprehensive overview of how Lorikeet is tackling the complex problem of building effective AI customer support agents, emphasizing practical innovation, human-AI synergy, and a deep commitment to understanding and meeting customer needs.
📝 About This Content
This article is based on insights shared by Teresa Torres on LinkedIn.
📅 Originally posted on May 28, 2026 | View original post on LinkedIn →