Your focus shifts from building to improving. You’ve launched with a solid foundation, but now the goal is to make your Agent smarter, more reliable, and more impactful. Improvement is about small, targeted changes that add up to big gains—whether that’s faster, more accurate answers, more relevant conversations, or smoother workflows.
Here’s a quick checklist to help you start improving your AI Agent right away, without reading through all the details first. If you’re comfortable jumping in without the full walkthrough, try this exercise in your own workspace:
Once your AI Agent is live, your focus naturally shifts from building to improving. You’ve launched with a solid foundation—but now it’s time to make it even smarter, more reliable, and more impactful. That’s where continuous improvement comes in.
The goal is to identify where your Agent is underperforming and apply focused enhancements that help it do more—and do it better. It’s not about starting over or rewriting everything. Instead, it’s about finding specific areas to improve: better responses, more relevant content, access to live data, or a more personalized experience.
Here’s how to identify gaps in your Agent’s performance, understand what’s going wrong, and choose the right tools to improve it:
Before you can improve anything, you need to know what’s not working as well as it could. Ada gives you two key tools to help with this: Topics and CSAT. Together, these tools provide a complete picture of both what’s going wrong and how it feels to the customer.
Prefer a conversational approach? Connect Ada to an AI assistant like Claude or ChatGPT using the MCP Server. Ask questions like “How can I improve my CSAT?” or “Analyze my unresolved conversations from last week” and get actionable recommendations without navigating dashboards.
Topics automatically group similar conversations based on end user intent. Each Topic surfaces key performance metrics that help you understand how well your AI Agent is handling a specific type of inquiry.
You can click on any Topic and drill down into the related conversations, making it easy to explore real examples where the Agent might be missing the mark. This helps you quickly spot patterns—like repeated Handoffs, vague responses, or missing Knowledge—and uncover concrete opportunities for improvement.
Focus on Topics that have high volume, low AR Rate, high AR Opportunity, and low CSAT. These signals indicate that:
In other words, these Topics represent high-friction, high-impact opportunities. Improving them can unlock quick wins—reducing Handoffs, improving satisfaction, and scaling automation where it matters most.
Want to see this in action? Check out the Refund requests and Card fulfillment examples. They show how to use Topic metrics and conversation patterns to identify issues and take targeted action—like adding content, refining messaging, or introducing automation—to improve performance and reduce Handoffs.
Customer Satisfaction (CSAT) surveys help you understand how end users feel about their experience — not just whether the AI Agent resolved their issue. A Conversation might be marked as resolved, but if the CSAT score is low, the answer may not have been clear, helpful, or empathetic enough.
To get the most out of CSAT data:
Want to see this in action? Explore the Account cancellation and Warranty replacement examples. They show how to use CSAT and Topic insights to spot friction, analyze related Conversations, and take focused action—like clarifying responses, asking better questions, or reducing unnecessary Handoffs.
Once you’ve identified a gap in your AI Agent’s performance—whether that’s from unresolved Topics, low CSAT, or customer feedback—the next step is choosing how to improve it. Ada gives you several tools, and each one is designed to solve a different kind of problem.
Here’s a quick guide to choosing the right one:
Each of these tools gives you a way to make your Agent better, smarter, and more helpful over time. The goal isn’t to fix everything at once. Start with the clearest gaps, apply the right kind of improvement, and measure the results. With each small change, you move your Agent closer to handling more on its own—with better outcomes for your team and your end users.
Let’s take a closer look at each one—what it does, why it matters, and how to apply it effectively.
Your AI Agent relies on your Knowledge base — the articles and resources it uses to answer questions. If that content is missing details or doesn’t cover common situations, the Agent might give poor answers or hand off too soon.
Use Knowledge improvements when:
You can improve outcomes by:
Want to see this in action? Check out the Payment troubleshooting and Refund guidance examples. They show how clearer, more focused Knowledge helps the Agent give better answers, reduce Handoffs, and improve the overall experience.
To perform at its best, your AI Agent needs access to more than just static content. That’s where Actions and Playbooks come in:
These tools improve your AI Agent’s performance by enabling it to:
Use Actions when static answers aren’t enough — like checking order status, verifying refund eligibility, or retrieving account info.
Want to see this in action? Check out the Card fulfillment and Subscription refund status examples. They show how automation can lead to faster resolutions, better experiences, and a more capable AI Agent.
Coaching helps you fine-tune how your AI Agent responds in specific situations — without changing its overall logic or retraining the model.
Use Coaching when the Agent gets the right intent but gives a response that’s:
With Coaching, you can make quick, targeted improvements — like updating a message, pointing to a better article, triggering an Action, or improving how the Agent escalates.
This helps your AI Agent:
Want to see this in action? Check out the Card fulfillment and Account cancellation examples. They show how small tweaks through Coaching can make a big impact on performance.
Personalization helps your AI Agent respond in a way that feels more relevant by using details like the customer’s location, language, membership tier, or order history.
Use it when the Agent gives correct answers, but they feel too generic or disconnected. Small changes — like mentioning a user’s plan or adjusting tone by region — can make responses feel more thoughtful and specific.
This helps your AI Agent:
Want to see this in action? Check out the Membership support and Product replacement by region examples. They show how using Personalization can make your Agent more helpful and engaging.