Customer: Whitepages
Whitepages is the premier provider of identity information, used by tens of millions of businesses and individuals each month to stay connected, prevent fraud, confirm shipping destinations, and make smarter real estate investments.
Whitepages innovates not only in the services it offers, but in how it builds and operates them — especially given the large volume of external requests to change or delete information. Using AI agentic automation, the Whitepages customer success team is reimagining how AI can transform complex operational backend workflows, improving response times, response quality, and productivity.
As an initial project, Whitepages focused on one specific type of customer success request — one that requires user verification, request validation, an automated update to internal systems, and customer confirmation. The volume of these tickets is roughly 500 per week.
Platform choice
Whitepages uses Zendesk as its ticketing system of record. When it came to AI automation, however, Whitepages explored handling automation outside Zendesk. The main factors in the decision were:
Reliability — How reliable is the AI automation system, and does it specialize in AI reliability?
Productivity — Can the system be used to rapidly create, test, iterate, and deploy agentic workflows?
Flexibility and power — Can it connect to other systems and handle a range of AI automation tasks?
Whitepages decided it was better to choose a best-of-breed AI agentic automation platform like Thunk.AI — one that could eventually automate a diverse range of operational workflows — rather than limit itself to the AI features within its SaaS ticketing product.
"Thunk.AI gives us the reliability, flexibility, and accuracy we need to automate sensitive customer workflows at scale. Its guardrails and no-code tools let us iterate quickly while maintaining full control. By launch, the automation was already surpassing our manual process. Ultimately, we needed an AI automation platform that could evolve with our operations, and Thunk.AI proved to be that partner."
— Nadine Thisselle, VP of Finance and Operations, Whitepages
Results of automation with Thunk.AI
This customer success workflow was built entirely on the Thunk.AI platform. As a no-code SaaS platform, Thunk.AI lowered the friction, time, and cost of creating and operating AI agents. The workflow was defined, owned, and operated by the customer success lead — the business process owner at Whitepages.
In just 3 months, with zero custom software development, Whitepages automated a critical, human-intensive customer success workflow using AI agents on the Thunk.AI platform.
The results in production:
50% reduction in response and resolution time for tickets
75% of tickets automated, with the rest intentionally redirected to human agents
90% decrease in cost per ticket processed
This successful automation allows Whitepages to be more effective in its use of customer success resources — engaging more deeply in the fewer cases that truly require a human touch.
Implementation details
The workflow has five steps, each defining AI agent logic in natural language. The overall workflow is orchestrated by the Thunk.AI platform, and the individual AI agents are hosted and executed in the Thunk.AI runtime. Every step relies on five foundational platform components for reliable execution:
AI Guardian — Mediates all AI agent invocations, checks and validates all responses, invokes tool calls, and validates tool-call results, providing "physical guardrails" against invalid behavior.
Schematized State — Every agent reads inputs from, and records results into, schematized state, providing "logical guardrails" against invalid behavior.
Tool Calling — Every agent has a configured set of tools (invoked via APIs). A broad set is available during development and is tightened significantly in production, improving reliability and predictability of outcomes.
Dynamic Reflection — Every agent can invoke dynamic planning and reflection to discover new tools, observe its progress, and course-correct toward the desired result.
Dynamic HITL — Beyond explicit human-in-the-loop settings, every agent can invoke HITL interactions when necessary (for example, when something is unclear or expected information is missing).
Thunk.AI lets teams choose among several LLMs, at both the overall-workflow level and the level of individual agentic steps. This implementation used GPT-4.1-mini — fast and inexpensive, while still delivering high quality.
Whitepages
To learn more about the various services offered by Whitepages:
Thunk.AI
To learn more about the Thunk.AI platform: