Our team has been setting up our Customer Agent to provide 24 hour support for customers and I was curious: what best practices your team have found?
Here are some of our findings:
Utilize a knowledge base to ground the chatbot's responses
Ensure the chatbot provides more accurate and relevant information by drawing from a curated set of documents
It's important to consider the format of the knowledge base documents
Pasting content into markdown or text files can help when PDFs can't be parsed
The knowledge base should contain relevant information such as industry facts, company data, and buyer beliefs
Regularly update and refine the knowledge base content
For assigning conversations to team members in HubSpot, ensure the correct settings are configured
This includes checking personal notification settings and assigning agents to chatflows
When building a knowledge base, consider how the language aligns with potential user queries
Including synonyms or common terms in the articles can increase the likelihood of relevant articles being retrieved
Our team found that the customer agent can struggle with semantics, for example even when provided a knowledge base titled "Pricing Document" prompts using the word "price" would return relevant answers but queries about "cost" went unanswered.
For knowledge base articles, consider including information that could answer potential follow-up questions
Including follow-up questions in the answers provided by your articles can help lead to more comprehensive and helpful chatbot responses
Consider a beta deployment of the chatbot to gather user feedback and identify knowledge gaps in real-world interactions
Once you have an initial knowledge base put together, go live! The customer agent's "Performance" tab has a "Knowledge Gaps" section that will give you feedback where your knowledge base is failing to answer questions
Continuously evaluate the chatbot's performance
Make it a habit to regularly check your conversations inbox to see how users are interacting with your agent
To keep your team looped in:
Send messages notifying your team when a new conversation comes in or is assigned to a human
HubSpot workflows offer slack/teams/google chat/ integrations as well as email sending
Explore options for handling scenarios where the bot cannot answer
Set up the rules in your chatflow for how the chatbot should handle transferring to a human, if your team has someone assigned to that role
Alternatively, if you cannot connect your users to a live representative of your team, enable email collection in your agent, allowing you to reach out and follow up with them
What best practices have you found to improve Customer agent performance?
Our team has been setting up our Customer Agent to provide 24 hour support for customers and I was curious: what best practices your team have found?
Here are some of our findings:
Utilize a knowledge base to ground the chatbot's responses
Ensure the chatbot provides more accurate and relevant information by drawing from a curated set of documents
It's important to consider the format of the knowledge base documents
Pasting content into markdown or text files can help when PDFs can't be parsed
The knowledge base should contain relevant information such as industry facts, company data, and buyer beliefs
Regularly update and refine the knowledge base content
For assigning conversations to team members in HubSpot, ensure the correct settings are configured
This includes checking personal notification settings and assigning agents to chatflows
When building a knowledge base, consider how the language aligns with potential user queries
Including synonyms or common terms in the articles can increase the likelihood of relevant articles being retrieved
Our team found that the customer agent can struggle with semantics, for example even when provided a knowledge base titled "Pricing Document" prompts using the word "price" would return relevant answers but queries about "cost" went unanswered.
For knowledge base articles, consider including information that could answer potential follow-up questions
Including follow-up questions in the answers provided by your articles can help lead to more comprehensive and helpful chatbot responses
Consider a beta deployment of the chatbot to gather user feedback and identify knowledge gaps in real-world interactions
Once you have an initial knowledge base put together, go live! The customer agent's "Performance" tab has a "Knowledge Gaps" section that will give you feedback where your knowledge base is failing to answer questions
Continuously evaluate the chatbot's performance
Make it a habit to regularly check your conversations inbox to see how users are interacting with your agent
To keep your team looped in:
Send messages notifying your team when a new conversation comes in or is assigned to a human
HubSpot workflows offer slack/teams/google chat/ integrations as well as email sending
Explore options for handling scenarios where the bot cannot answer
Set up the rules in your chatflow for how the chatbot should handle transferring to a human, if your team has someone assigned to that role
Alternatively, if you cannot connect your users to a live representative of your team, enable email collection in your agent, allowing you to reach out and follow up with them
What best practices have you found to improve Customer agent performance?
In addition to your practices, refine NLP models for better understanding, set up clear human handoff triggers, proactively engage users, and integrate multi-channel support. Gather feedback through surveys, ensure context retention during handoffs, and track success metrics like FRT and CSAT for continuous improvement in chatbot performance.
Our team has been setting up our Customer Agent to provide 24 hour support for customers and I was curious: what best practices your team have found?
Here are some of our findings:
Utilize a knowledge base to ground the chatbot's responses
Ensure the chatbot provides more accurate and relevant information by drawing from a curated set of documents
It's important to consider the format of the knowledge base documents
Pasting content into markdown or text files can help when PDFs can't be parsed
The knowledge base should contain relevant information such as industry facts, company data, and buyer beliefs
Regularly update and refine the knowledge base content
For assigning conversations to team members in HubSpot, ensure the correct settings are configured
This includes checking personal notification settings and assigning agents to chatflows
When building a knowledge base, consider how the language aligns with potential user queries
Including synonyms or common terms in the articles can increase the likelihood of relevant articles being retrieved
Our team found that the customer agent can struggle with semantics, for example even when provided a knowledge base titled "Pricing Document" prompts using the word "price" would return relevant answers but queries about "cost" went unanswered.
For knowledge base articles, consider including information that could answer potential follow-up questions
Including follow-up questions in the answers provided by your articles can help lead to more comprehensive and helpful chatbot responses
Consider a beta deployment of the chatbot to gather user feedback and identify knowledge gaps in real-world interactions
Once you have an initial knowledge base put together, go live! The customer agent's "Performance" tab has a "Knowledge Gaps" section that will give you feedback where your knowledge base is failing to answer questions
Continuously evaluate the chatbot's performance
Make it a habit to regularly check your conversations inbox to see how users are interacting with your agent
To keep your team looped in:
Send messages notifying your team when a new conversation comes in or is assigned to a human
HubSpot workflows offer slack/teams/google chat/ integrations as well as email sending
Explore options for handling scenarios where the bot cannot answer
Set up the rules in your chatflow for how the chatbot should handle transferring to a human, if your team has someone assigned to that role
Alternatively, if you cannot connect your users to a live representative of your team, enable email collection in your agent, allowing you to reach out and follow up with them
What best practices have you found to improve Customer agent performance?
In addition to your practices, refine NLP models for better understanding, set up clear human handoff triggers, proactively engage users, and integrate multi-channel support. Gather feedback through surveys, ensure context retention during handoffs, and track success metrics like FRT and CSAT for continuous improvement in chatbot performance.
Thanks so much for sharing these great best practices! They’re really helpful. @TomM2 and @Lucila-Andimol curious if you’ve had any additional tips or experiences with improving Customer Agent performance?
.png "HubSpot Employee"){kind=icon}
