HOW WE BUILT “BETH”: OUR 90-DAY JOURNEY TO TRANSFORM BENEFITS ADMINISTRATION WITH AI

Let’s be honest—when we first started talking about building an AI agent to help our benefits team, it felt a bit like science fiction. But after watching our senior benefits specialists spend countless hours digging through thousands of documents during open enrollment season, we knew something had to change. What we ended up with was “Beth,” an AI agent that transformed how we serve our tens of thousands of worksite employees. Here’s our story.

At Questco, our mission is to be the greatest ally to the growing business. And we accomplish this through human-centered, expertise-driven service. We’ve built our reputation on delivering comprehensive PEO solutions that allow business owners to focus on what they do best while we handle the complexities of HR, benefits, and compliance.

Our technology mission reflects this same commitment: “Empower Questco’s growth by delivering best-in-class technology platforms and services that enable success for both our employees and our customers. Unleashing greatness at every step.”

But here’s the thing about applying AI in a PEO—it’s all about balance. We’ve learned that successful AI implementation isn’t about replacing the human experience our customers value. Instead, it’s about using AI to optimize our internal operations so our team can spend more time providing that personal touch our clients depend on. Think of it as cost-to-serve optimization that actually enhances rather than diminishes the human experience.

THE PROBLEM THAT KEPT US UP AT NIGHT

If you’ve ever been through open enrollment at a PEO, you know it can feel like organized chaos. Picture this: tens of thousands of worksite employees across hundreds of client companies, each with their own unique benefit packages, carrier requirements, and deadlines. Now imagine our senior benefits team trying to navigate through thousands of internal documents to answer questions like “What’s the exact enrollment deadline for Company XYZ’s dental plan?” or “Which carriers accept late enrollments for this specific client?”

Our team was drowning in paperwork. We had this massive internal library that we kept meticulously updated with benefits processes, carrier information, and client-specific materials. The challenge wasn’t that we didn’t have the information—we had plenty of it. The problem was finding the right piece of information quickly when our phones were ringing off the hook and deadlines were looming.

During peak open enrollment, our senior benefits specialists were spending way too much time searching through documents instead of doing what they do best: helping our clients make informed decisions about their benefits. It was frustrating for everyone involved, and frankly, it wasn’t the level of service our clients deserved.

WHAT WE CONSIDERED AND WHY WE CHOSE OUR PATH

When we sat down to solve this problem, we looked at three main options:

Option 1: Scale up with more staff. We could hire additional temporary specialists during open enrollment. The math was pretty sobering: over $300,000 annually when you factor in recruitment, training, and the premium rates for qualified temporary benefits experts.

Option 2: Build a proprietary AI solution from scratch. This would have been the “Cadillac” approach, building a custom AI platform from the ground up. Price tag? Somewhere between $200,000 and $400,000, with a timeline that would have taken us well past the next open enrollment season.

Option 3: Leverage our internal platform and extend it with a simple AI agent. This was our “let’s be smart about this” option. We could build on our existing technology foundation and add about $50,000 in additional support to create an AI agent focused specifically on document search and retrieval.

We chose Option 3, and honestly, it was one of the best decisions we’ve made. We called our AI agent “Beth” (because every good AI needs a name), and she became the teammate our benefits staff never knew they needed.

BUILDING BETH: THE 90-DAY SPRINT

What made this project exciting was the timeline we set for ourselves: 90 days from concept to launch, with several iterations afterward to refine and expand Beth’s capabilities. Looking back, it feels like we packed about six months of work into three months, but the urgency kept us focused on what really mattered.

The First 30 Days: Foundation and Planning
We started by getting our cross-functional team together—benefits experts, IT folks, and key stakeholders who would actually be using Beth. The biggest lesson we learned early on was that this couldn’t be an “AI for AI’s sake” project. We had to stay laser-focused on solving the specific problem: helping our team find the right information fast.

Data preparation was huge during this phase. We discovered that about 70% of AI project failures come from poor data quality, so we spent serious time organizing and cleaning up our document library. It wasn’t glamorous work, but it was essential.

Days 31-60: Rapid Development
This is where things got really interesting. Instead of traditional development sprints, we used what’s called an AI-Driven Development Lifecycle with intensive “bolts” of focused work. We’d build something, test it immediately, get feedback from our benefits team, and iterate quickly.

Beth started simple—basically a smart search engine that could understand natural language questions about our benefits documents. But we could see the potential almost immediately. When someone asked “Can I make changes to my FSA contributions” Beth could pull the exact information from our compliance documents in seconds.

Days 61-90: Launch and Initial Deployment
The final month was all about getting Beth ready for prime time. We did extensive user training with our benefits team, ran pilot tests with a small group, and then rolled out the full deployment with monitoring systems in place.

Our target was heavy user adoption and quantifiable time savings, and honestly, we exceeded that. Once our team saw what Beth could do, they were hooked.

WHAT BETH ACTUALLY DELIVERED

The results have been impressive so far. Beth has already freed up significant time from our benefits staff that we’re repurposing to provide even better service to our clients. Instead of spending 20-30 minutes searching through documents to answer a complex benefits question, our team can now get that information in under a minute and spend the saved time actually talking with clients about their options.

We’ve seen error rates drop and response times improve dramatically in our day-to-day operations. More importantly, our benefits team is already enjoying their work more because they can focus on the human aspects of their job rather than document hunting.

But the real test is coming this October through January during open enrollment season. That’s when we’ll see Beth’s true value. While other PEOs will likely be scrambling with temporary staff and longer response times, we’re confident our team will operate at peak efficiency. Beth will handle the routine information retrieval during those crazy busy months, allowing our experts to focus on strategy, problem-solving, and providing the personalized guidance our clients need when making critical benefits decisions.

THE PATH AHEAD: WHAT’S NEXT FOR AI AT QUESTCO

Beth was a great experiment to show us the art of the possible. We’re actively looking at how AI can unlock opportunities for our employees and clients in the future. The key is maintaining that balance—using AI to enhance human capabilities rather than replace them. We look forward to the future this exciting technology will bring and the value it will create for our clients through the expertise and human interaction that’s central to our business.

LESSONS WE’D SHARE WITH OTHER PEOS

If you’re considering AI implementation, here’s our honest advice:

Start small and focused. Don’t try to build the perfect AI solution right out of the gate. Pick one specific problem that’s causing real pain and solve that first. Beth wasn’t revolutionary technology—she was the right solution to a specific problem.

Get your data house in order. This can’t be emphasized enough. Your AI is only as good as the information it has access to. Invest the time upfront to organize and clean your data.

Plan for iteration. Beth got better every month after launch because we kept refining her based on real-world usage. AI systems improve through use, so don’t expect perfection on day one.

Keep humans in the loop. The goal isn’t to replace your expertise—it’s to amplify it. Our benefits team is still making the complex decisions and providing the personal touch. Beth just makes them more efficient.

Set realistic timelines. Ninety days was aggressive, but it worked because we stayed focused and had strong executive support. Don’t rush the foundation work, but don’t get stuck in planning mode either.

THE BOTTOM LINE

Building Beth taught us that AI doesn’t have to be intimidating or expensive to be transformative. Sometimes the best solution is the simplest one that directly addresses your biggest pain point. For us, that was helping our team find information faster so they could spend more time doing what they do best—serving our clients.

The PEO industry is evolving rapidly, and technology will play a crucial role in how we differentiate ourselves. But at the end of the day, it’s still about people helping people. AI just gives us better tools to do that more effectively.

Beth has become an integral part of our team, and we’re excited to see how she performs during our upcoming open enrollment season. If you’re thinking about AI implementation at your PEO, our advice is simple: start with a real problem, keep it focused, and don’t be afraid to iterate. You might just build something that transforms how your team works.