Explorer: The agent-optimizing agent
If you've ever built an AI agent, you know the hard part isn't getting to launch. It's the continuous work of understanding how your agent performs, and improving it, that comes after.
As it turns out, the best way to improve AI is with more AI. Explorer, the agent-optimizing agent, works alongside Ghostwriter, the agent-building agent, to proactively tell you what needs fixing or improving, and how. Think of it like ChatGPT deep research, but instead of doing research on the Internet, it's doing research over your customer conversations. Since its launch last year, hundres businesses have used Explorer each week, including leading brands like ADT and DIRECTV.
Diagnose what's driving performance
Metrics tell you what's happening. Explorer tells you why by prompting you with questions you may not have thought to ask, like:
- What are customers frustrated about that the agent isn't resolving?
- We released a new subscription plan last week — what do members find confusing about it?
- Cancellations are up in the past week —what changed?
A global activewear brand was seeing NPS dip but couldn't pinpoint why. In one Explorer session, they discovered their agent was handling a key customer interaction too aggressively — a nuance buried across thousands of conversations that would never have surfaced in a dashboard. Explorer identified it and recommended exactly how to fix it. They adjusted the experience, and NPS improved.
“At ADT's scale of hundreds of thousands of conversations, manual QA is impossible. The deep research in Explorer has been a massive step forward for ADT’s Sierra Platform, allowing us to uncover hard to analyze performance vectors that were previously invisible. By combining Virtual Agent Performance with Customer Experience insights, we’ve moved beyond just identifying problems to strategically driving growth in both containment and customer experience.”
Matt Robbins
Manager Intelligent Automation
Test a hypothesis before you build
Improving your customer experience requires continuous experimentation. Explorer helps you double down on your best ideas by identifying what's working and why. And when you want to go deeper, you can ask questions like:
- We launched an experimental order return flow last week — should we roll it out to more customers?
- We're thinking of improving our onboarding process. Which customer segment should we focus on first?
- We're seeing more troubleshooting cases — is that specific to mobile or is it across all channels?
The customer experience team at a national homebuilding company suspected customers were dropping off because the agent wasn't personalizing responses based on where they were in the buying journey. Before investing resources to make the change, they brought that hypothesis directly to Explorer — asking it to analyze conversations across customer segments and buying stages. Within a single session, they had enough signal to know exactly which parts of their hypothesis held up and which didn't. They went into their next sprint with confidence, not guesswork.
“Explorer is transforming how we understand and optimize customer interactions. Our teams can ask plain‑English questions and quickly see what’s happening, where to focus, and what actions will have the biggest impact without digging through data. That faster access to insight helps us make better decisions. It’s a powerful step forward in making DIRECTV’s customer experience smarter, faster, and more data-driven at scale.”
Ryan Mann
AVP of Digital Services
Anticipate what’s coming
The most valuable insight isn't always about what happened. Sometimes it's about what's going to happen next. We see teams preparing for what's ahead by asking Explorer questions like:
- What did customers ask for last season that we weren't ready for?
- Based on traffic this past year, what customer questions should we expect to spike during our holiday season?
- What capabilities should we add to our agent to most improve our customer satisfaction?
When a leading health insurer approached open enrollment season, they didn't wait for call volume to spike. They used Explorer to surface what customers had asked the prior year, taking those insights to automatically update their agent's knowledge base before enrollment opened, and reducing transfers from day one.
Always on, always improving
Explorer already knows what to look for. Beyond the questions you bring to it for deep analysis, it runs continuously in the background — scanning every customer conversation so the most important signals always reach you. Every week it delivers a briefing with what's changed, what's causing it, and recommendations you can implement in one click with Ghostwriter.
Closing the continuous improvement loop
The old model was understand, prioritize, build, wait. Explorer and Ghostwriter collapse that into a single loop — one that turns every customer conversation into a better agent experience, continuously and automatically. This is what it looks like when AI optimizes AI.
