Client story Kentucky Department of Education targets cutting teacher certification call backlog by 20% with AI

Business challenge

KDE faced a major hurdle tied to a nationwide trend: a serious challenge in finding and keeping teachers. New teachers, or professionals looking to change careers into education, often need clear answers about certification. This search for answers created a massive problem for the KDE team.

Waiting for answers to simple questions

“Over 50% of the call volume that comes into the Kentucky Department of Education is targeted at our Office of Educator Licensure and Effectiveness,” says Park. “Historically, these questions have needed people to answer them. We work extremely hard to be as responsive as possible, and during peak times our team members are stretched thin while handling a high volume of calls.” 

This high call volume means people may have to wait for answers. Team members can spend their days answering simple, repetitive questions instead of focusing on the hard, personal scenarios where their expertise is needed most.

Relaunching the website as a single source of truth

After recently cleaning up and relaunching their prospective teacher website, GoTeachKY.com, KDE now has a reliable single source of truth for information on how to become a teacher in Kentucky. Their challenge wasn’t the quality of the information, but how to get it to more people faster. 

To solve the challenge, KDE decided they needed a technology solution that would relieve their team from answering basic questions and improve public service.

Solution

KDE set out to find a solution that met three key standards: it had to come from a solution provider they trusted, fit within their budget, and use existing, ready-to-configure tools instead of custom software built from scratch.

“For this project, we wanted to work with our trusted solution providers that we already have contracts with,” says Park. KDE chose to use Google Cloud for the platform and Insight, an Insight company and multiple-time Google Cloud Partner of the Year, for the implementation. 

Building and deploying the AI agent quickly

Insight quickly built and deployed an AI agent, which KDE named “Diego” (Ask Diego), using Google Cloud’s Vertex AI. This let the team create a specialized agent without the time and cost of a custom build. The solution used a few key technologies to smooth out the process:

• Cloud Conversational Agents (Dialogflow CX) served as the core engine for building and fine-tuning the intelligent agent.
• Google Dialogflow Messenger let Insight integrate the agent chat window right into the website, giving users an easy-to-use interface.
• A Google Vertex AI data store regularly indexed the website content, ensuring that Diego always had the freshest, most current information.

KDE and Insight used a “walled garden” approach, so Diego only searched the information on the department’s website, which kept the answers safe, secure, and responsible. Insight also built guardrails and instructions into the agent to make sure Diego gives correct answers.

To maximize the value for the KDE team, Insight offered a flexible approach, accelerating the pace at which KDE was comfortable and slowing down to give extra support when it was needed. “Insight quickly shifted gears when we needed them to,” says Park. “For example, they would back up and focus on our current questions instead of covering scheduled content. This hands-on, adaptive approach ensured that our team learned how to keep the solution running on their own.”

Impact

The implementation of the AI agent created quick wins and big wins for KDE, immediately transforming how they serve customers and support their team.

Quantifiable benefits

KDE has seen clear signs that Diego will meet their main goals of reducing calls and improving customer service.

• Service availability: The agent gives answers 24/7, which is a big deal for people asking questions after hours.
• Call support offset: The agent is fulfilling the initial goals while targeting a cut of up to 20% of the call volume for basic questions, which means those questions never reach a team member.
• Low cost: The use of existing, configurable Google Cloud tools meant the project stayed within budget. “So far, the agent spend has been very small, proving the project’s value when compared to the cost of hiring new staff,” says Park.
• Deep customer engagement: The agent is capable of giving more than one quick answer at a time. KDE monitors agent conversation threads and sees high engagement, with some users asking 10 to 11 follow-up questions in a single session. This shows that the agent can handle deep, complex questions. “One conversation ended with the user saying, ‘Thank you so much’ after Diego answered detailed questions about certification application fees and processing time,” says Park.

Qualitative improvements

The solution changed the way the KDE team works and how they think about their published content.

• Improved quality for internal teams: Diego’s value goes beyond public service. Internal teams use the agent to test their own website. “It’s not just for the public customer; the AI agent helps us check if answers on the website are incorrect and why. That helps us get the correct content online,” says Park. “This turns the agent into a quality control tool that improves the department’s information.”
• Team skill elevation: Working on the AI agent project gave the KDE team a professional learning opportunity. They learned how to build and support a customer-facing AI tool, expanding their skills and getting fired up about the new technology.
• Better customer service: The agent helps KDE avoid losing potential teachers who might have given up because of a slow response. “If we can service them faster, answer their questions better, and offer an on-ramp, we can offset up to 20% of basic questions with the Google AI agent,” says Park. “Then we can spend more one-on-one time with candidates who have very specific questions that can only be answered by our people with deep expertise.”

Thinking differently about what’s possible

The AI agent’s quick success changed what KDE thought was possible. The team’s confidence has grown, and they are already working with Insight on other projects.

“Artificial intelligence won’t replace teachers, and it won’t replace our employees,” says Park. “It’s only going to help us work better and improve our resources, services, and tools.”

KDE is now looking at where else they can place similar AI agents. They are also exploring how to use Google Cloud’s Vertex AI and Gemini for other projects. The positive experience of this first project caused the team to think differently. It changed what they feel is possible, leading to new ideas for continued growth and innovation.