Gladly AI Conversations Page Overview

Prev Next

The Gladly AI Conversations page contains all recent Customer interactions with Gladly AI, detailing its responses and decision-making process for answering questions or deciding when to hand off to Agents. The Gladly AI Conversations page corresponds with Guides and Journeys pages – together detailing how Gladly has been configured to respond to and engage with Customers.

The Gladly AI Conversations page is only visible to certain permission sets

This page can be accessed by Gladly users with Team Member or Admin permissions. It is not visible to Agents.

Find Gladly AI Conversations

The Gladly AI Conversations page lists recent Conversations that Gladly has had with Customers and provides real-time updates on how Gladly is engaging. Use Search and Filters to find recent Conversations where Gladly engaged.

Gladly AI Conversation history is preserved for 30 days

When searching for historical Conversations within the Gladly AI Conversations view, keep in mind that Conversations are only available for 30 days after the Conversation was first created.

Overview of Gladly AI conversations with status and outcome details for each user.

You can filter the list of recent Gladly interactions using the following filters:

  • Status – Shows whether a conversation has already been closed or whether it’s still in progress. Possible values are In Progress or Closed.

  • Gladly Outcome – Once Gladly has completed its work on a Conversation, this filter will show the outcome of its interactions with the Customer. Possible values are Gladly Did Not Respond, Gladly Handed Off, and Gladly Resolved. Here’s what those values mean:

    • Gladly Did Not Respond = Gladly transferred the conversation without sending a Generative AI response to the consumer. For example, the Customer types “Agent” before Gladly sends a response generated by AI, and Gladly immediately transfers to an Agent. These Conversations are not counted as Billable Gladly AI Assisted Conversations.

    • Gladly Handed Off = Gladly responded to the consumer and handed off to an Agent without resolving the conversation.

    • Gladly Resolved = Gladly responded to the consumer and did not hand off to an Agent during the course of the conversation. This Gladly Outcome is only applied to a conversation after it has been closed.

  • Type – Shows whether Gladly responded to the Customer using an Answer, based on information contained in your company’s Public Answers, or with an Action, based on information contained in your company’s Guides. Possible values are “Answers” or “Actions.”

  • Channel – Shows the Channel on which the Conversation took place. Possible values are “Chat,” “SMS,” or “Phone Call.”

Example use cases

  • Resolved Conversations

    • Filter the Conversation list using the “Gladly Resolved” outcome type and review Conversations where Gladly has resolved the Customer issue without needing to hand off to an Agent.

  • Handed off Conversations

    • Filter the Conversation list using the “Gladly handed off” outcome type and review Conversations where Gladly started to help the Customer but handed off to an Agent to resolve the inquiry.

  • Conversations where Gladly did not respond

    • Filter the Conversation list using the “Gladly did not respond” outcome type and review Conversations where Gladly did not respond to the Customer with an AI-generated response and instead handed off to an Agent immediately.

Interpret Gladly AI Conversation details

Click into any Gladly interaction from the Gladly AI Conversation list to see the details of the Customer inquiry, how Gladly responded, and why Gladly decided to respond to the Customer or to hand off to an Agent to resolve the Customer’s inquiry. Review Manage Gladly AI Conversations Page to understand how to leverage components of the Gladly AI Conversation page for ongoing QA.

Conversation Status

Every Customer interaction is accessible on the Gladly AI Conversations page, with real-time updates and statuses that provide insight into how Gladly engages with Customers.

Below is a table of possible Conversation statuses.

Status

Description

Gladly resolved, Closed

Gladly resolved the Customer’s questions without needing to hand off to an Agent, and the Conversation has ended.

Gladly handed off, In Progress

Gladly engaged with the Customer but had to hand off the Conversation to an Agent, and the Conversation is still in progress.

Gladly handed off, Closed

Gladly engaged with the Customer but had to hand off the Conversation to an Agent, and the Conversation ended.

Gladly did not respond, In Progress

Gladly could not provide an initial response that passed the quality check, so it provided a summary and handed off to an Agent. The Conversation is still in progress.

Gladly did not respond, Closed

Gladly could not provide an initial response that passed the quality check, so it provided a summary and handed off to an Agent. The Conversation has ended.

Conversation Review

The Conversation Review is a powerful feature that provides complete transparency into how Gladly generates a response or decides to hand off a Conversation to an Agent.

This review appears within the Gladly AI Conversation page every time Gladly, acting as a "behind-the-scenes", look at its decision-making process.

Use the Conversation Review to:

  • Gain Clarity: See exactly which Guides, actions, rules, and guidance were used to generate Gladly’s response.

  • Improve Quality: For handed-off Conversations, understand precisely why Gladly’s reply failed specific quality checks. This insight is key to refining your Guides and reducing future quality check failures.

Access the Conversation Review

  1. Click on the top left corner of the screen.

  2. Select Gladly AI Conversations.

  3. Select a Conversation where Gladly has engaged.

  4. Click next to Gladly sent a chat to expand the Conversation Review.

  5. Expand sections within the Conversation Review for more details as needed.

Conversation Review items are available for each message that Gladly engages with

Click Review Previous Message to go to the prior Conversation Review box.

Conversation Review Components

The review is structured into distinct, easy-to-follow sections. Here’s a breakdown of what each section tells you.

Guidance on verifying customer information and updating shipping address in a conversation.

[A] Review Previous Message

This section allows you to view the previous Conversation Review box for the previous message that Gladly sent to the Customer.

[B] Guide Used

This showcases the Guide that Gladly leveraged to answer the Customer’s inquiry. Click on the Guide name to review its contents.

[C] Section Used and Rationale

This portion of the Conversation Review shows the section(s) Gladly used to work through the Customer’s request. It also displays the various action(s), rule(s), and pieces of guidance that Gladly incorporated into its response.

[D] Actions, Evaluated Guide Rules & Guidance Used

This is where Gladly evaluates the items configured within the section. This could include, any selected actions, conditional logic you’ve set up as rules, and individual pieces of guidance.

  • Action: This shows the specific action Gladly took within a given Guide section.

    • Example: “Verify Full Name”

  • Evaluated Rules: These are the specific conditions Gladly checked against. It will list the rules and whether or not they were met.

    • Example: “full name is verified” or “full name is not verified.”

  • Guidance Used: This is the resulting instruction Gladly followed based on the evaluated rules.

    • Example: "If the full name is verified, then go to ask for new shipping address."

[E] Quality Check Results

This part of the Conversation Review shows whether Gladly’s proposed response passed or failed the quality check.

  • If the response passed, no issues will be flagged.

  • If the response failed, Gladly hands off to an Agent, and this section details the precise quality checks that were missed and the rationale.

When the quality check has failed, the Conversation Review will display the following:

  • Proposed Response

    • This is the message Gladly would have sent to the Customer, if it had passed the quality check.

  • The specific quality checks that have been flagged

    • These flags explain why the proposed response was blocked.

    • Example: Missing action or Unverified information

[F] Gladly Hand Off

If Gladly hands the Conversation off to an Agent, the Conversation Review will indicate the handoff reason and provide a rationale. Handoff reasons can be any of the following:

  • A failed quality check

    • “Gladly handed off the conversation due to a failed quality check.”

  • Guidance built into the Guide that instigated a handoff

    • “The section guidance led Gladly to perform a handoff with the reason [your section guidance].”

  • An action built into the Guide hit an error when run

    • “Something didn’t go as planned when running the Action [Action name].”

  • An error occurred when running the Guide

    • “Something didn’t go as planned when running the Guide [Guide name].”

How to Use the Conversation Review

Use the Conversation Review to:

  • Audit Performance: Check if Gladly is correctly interpreting and executing the actions, rules and/or guidance you’ve configured in your Guides.

  • Refine Guides: Review hand off rationales to understand why Gladly passed the Conversation on to an Agent and identify where to update your Guide to better align Gladly’s actions with its language.

  • Ensure Quality: Gain confidence that Gladly’s built-in guardrails are working to prevent inaccurate or incomplete information from reaching your Customers.

Quality Checks

Every response that Gladly generates undergoes automated quality checks before reaching the Customer. These quality checks run in the background and act as guardrails to prevent Gladly from sending incomplete, inaccurate, or problematic responses.

Quality check results are visible to Admins and Team Managers within individual Conversations in the Gladly AI Conversations page. The information displayed in the Conversation Review will vary depending on whether Gladly’s response passed or failed the quality checks.

How Quality Checks Work

Quality checks use a combination of AI-powered algorithms and code-based validations to evaluate Gladly’s proposed responses across three key areas:

Category

What It Checks

Missing Action

Flags claims where Gladly states it will perform or has performed actions not supported by

your configuration

Missing Information

Flags when Gladly lacks adequate information to respond appropriately to the Customer's

situation

Unverified Information

Flags unverified information like dates, policies, contact details, or facts not found in available

knowledge sources

When a quality check fails, Gladly automatically hands off the conversation to an Agent and provides context about why the response was blocked.

Learn more about quality checks

For detailed information about specific quality check algorithms, how they're configured, and best practices for reducing quality check failures, see Quality Checks Overview.

Handoff to Agent

Gladly always indicates a hand off to an Agent with hand off messaging [B], if the proposed response fails the quality check [A]. Gladly will generate a Handoff summary [C] for the Agent to provide context around the automated exchange prior to the hand off.

Chat conversation discussing reward points and customer assistance details.

  • Gladly will always hand off to an Agents whenever it cannot provide a proper initial or subsequent response.

  • The handoff message sent to a Customer during a handoff [C] shown as “Hold tight! Calling in one of our Brand Ambassadors to help” is customizable within the “When to get more help” tile in the Gladly Guides interface.