Building your best Customer Agent
The most effective Customer Agents are built with intention, maintained consistently, and improved over time. These best practices can help you focus on the highest-value customer needs, expand agent capabilities thoughtfully, and use performance insights to drive ongoing optimization.
BEST PRACTICE #1Start with the highest-value customer questions and journeys
Ensure your Customer Agent is prepared to handle what customers ask the most.
Map your customer journey
Put yourself in your customers’ shoes. What does a typical journey with your brand look like? What common questions arise? Where might they need a little extra support from your agent?
Ensure your storefront is dynamically syncing
Your Customer Agent should respond and align with what your customers are seeing online. Enable the Turn on auto-syncing setting so that any changes to your storefront are reflected in your Customer Agent’s knowledge.
Fill in gaps with high-priority additional content
If there are other frequently asked questions or high-priority items that aren’t clearly covered by your site, add those first as separate pieces of content so your Customer Agent can answer questions on those topics effectively. You can upload formats such as PDF, .docx, .txt, etc.
Train and govern the agent continuously, not just once
A truly effective Customer Agent isn’t just a set it and forget it tool. It’s continuously improved over time to always meet the needs of your customers.
Perform a weekly audit of conversations
Review how the agent handles conversations and escalations on a weekly basis. When is it getting content wrong? Where is it sounding off-brand? When are escalations mishandled?
Refine and add content and governance
Where there’s missing or incorrect information, refine or add supplemental content to your Agent. For tone and language issues, create a more detailed custom tone and set more specific communication rules to align closer with your brand voice. When handoffs aren’t happening as expected, improve your escalation logic by refining escalation rules and adding additional custom skills.
Preview and test before setting live
Before setting any improvements and adjustments live, preview your agent to ensure your tweaks are reflected as expected in Customer Agent responses. Agent Builder can help you continue making iterations if need be.
Expand capabilities where the agent can create clear customer and business value
How can your Customer Agent do more? Grow your service possibilities with custom skills and more channels.
Create custom skills for high-value use cases
Build new custom skills to empower Customer Agent to help resolve issues that aren’t covered by out-of-the-box skills. Focus on recurring and high-value issues to drive the most value.
Use Agent Builder to create skills with natural language
Creating custom skills is a perfect use case for Agent Builder. Describe what you want the skill to do, how it should behave, and when it should handoff, and Agent Builder will generate a full draft for your review and publish.
Launch across new channels
Meet your customers where they are by launching Customer Agent across web chat, email, text message, and WhatsApp. If you have Customer Hub, host your Customer Agent in your hub to make getting support easier than ever.
Review performance regularly and treat gaps as optimization opportunities
Improving your Customer Agent over time empowers you to consistently meet the needs of your customers. Review performance and implement changes regularly with the below strategies.
Start with the Performance dashboard
Make the Performance dashboard your home base for evaluating Customer Agent because it shows overall performance, post-conversation customer actions, business impact, and how volume is split across skills. Use it first to spot which trends or metrics need a closer look.
Use deeper views to diagnose what’s driving results
Review Analytics > Metrics when you want to track specific Customer Agent events over time, and check the profile activity feed when you need to understand an individual shopper experience. Together, these views help you move from high-level performance trends to specific causes.
Run simulations before making major changes
Test updates in simulations to see how the agent responds before you roll them out live. This helps you catch weak guidance, missing content, or skill issues early.