# Chat quality review board for LiveChat support leads

> Score last week's finished live chats against one rubric, let an AI grader draft the notes with quotes, and send coaching straight to the agent.

- Workflow type: app
- Services: LiveChat, Slack, General Input Database
- Categories: Customer Support, Operations
- Published: 2026-08-31

## What it does

- Pulls last week's finished live chats into one review queue, with badly rated and unusually long conversations floated to the top so your leads start where the problems are
- Shows the full conversation next to a four part scorecard covering greeting, diagnosis, tone and resolution, so reviewers read and score on the same screen
- Assigns each chat to one reviewer, so two people never quietly score the same conversation and nobody gets reviewed twice
- Grades a chat for you on request: an assistant reads the transcript, scores it against your rubric, quotes the exact lines behind every score and drafts coaching notes your reviewer can accept or change
- Applies a review tag back onto the conversation once a review is accepted, and can send the coaching note to the agent as a private Slack message
- Trends each agent's average review score over time next to their live chat rating, so you can see whether last month's coaching actually moved the number

## What you'll need

- A LiveChat account with chat history you can read, plus permission to add tags to conversations
- A team lead or reviewer login for each person who will score chats
- A Slack workspace, only if you want coaching notes delivered as private messages to your agents
- An agreed scoring rubric; the app starts with greeting, diagnosis, tone and resolution, and you can rename or reweight these

## Prompt

I want a chat quality review board where my support leads work through last week's finished LiveChat conversations, instead of spot checking inside the LiveChat archive. It is an internal tool for a small group of reviewers, not something my frontline agents open.

The main view is a review queue of ended chats. Load them with the LiveChat List Archives operation, defaulting to the last seven days, and show one row per conversation with the date, the agent who handled it, the customer, the chat rating, its tags and the duration. Let reviewers filter by rating, by tag, by agent and by duration. Populate the tag filter from List Tags and the agent filter from List Agents so people pick from real values rather than typing them. Sort the queue so the chats most worth reviewing float to the top: badly rated chats first, then unusually long chats, where unusual means well above the median duration for the current result set. Show each row's review state in the queue, so it is obvious what is unreviewed, claimed, graded by the assistant but not yet accepted, and finished.

Opening a row shows the conversation in full on one side and a scoring form on the other. Build the transcript from Get Chat for the conversation, and use List Threads to pull every thread in that chat so a conversation that was resumed reads as one continuous history rather than a single fragment. Render it as a readable back and forth with speaker names and timestamps, clearly separating customer messages from agent messages, and include the rating and any comment the customer left. The scoring form covers four areas: greeting, diagnosis, tone and resolution. Each gets a score and an optional comment, plus one overall note for the chat.

Store reviews in the General Input Database, one record per chat per reviewer, holding the chat and thread identifiers, the agent being reviewed, the four scores, the comments, who reviewed it, when, and the current state. Reviewing is claimed work: when someone opens an unreviewed chat to score it, claim it for them, and show other reviewers that it is taken so two people never review the same chat twice. A reviewer can release a chat they claimed but did not finish. Never lose a partly filled form, and let a reviewer come back and edit their own review before they accept it.

Add a "Grade this chat" button that kicks off a background agent for that conversation. The agent reads the full transcript, scores it against the exact same four part rubric the reviewer sees, and for every score quotes the specific customer and agent lines it based that score on, so a reviewer can check the reasoning without rereading the whole chat. It also writes short coaching notes aimed at the agent who handled the conversation: what went well, what to do differently next time. Write all of that back into the app against the chat, so the review screen shows the assistant's score and quotes beside the reviewer's own form. The reviewer can accept the assistant's scores as their own, override any individual score, or rewrite the coaching note. Make it clear on screen which scores came from the assistant and which the human set, and keep both after an override so we can tell later how often the grader agreed with reviewers. Grading one chat should not block the queue, so let a reviewer kick it off and keep working while it runs.

Accepting a review finalizes it and applies a tag back to the conversation in LiveChat with Tag Thread, using the thread the review was written against. Which tag depends on the outcome: something like qa-reviewed when the chat passes and qa-failed when it falls below the passing score, with both tag names configurable. Optionally also write the verdict onto the conversation with Update Thread Properties, so the result is visible to anyone working inside LiveChat and not only in this app. If a tag write fails, keep the review saved and show the failure with a retry rather than silently dropping it.

On accepting, the reviewer can also send the coaching note to the agent who handled the chat as a Slack direct message. Use Open a Conversation to open the direct message channel with that agent, then Send a Message to deliver the note. Send it as the reviewer rather than as a bot, since this is one person coaching another. Map LiveChat agents to Slack users by email, let the reviewer confirm or correct the recipient before anything sends, and always show a preview of the message. Sending is optional per review, and record on the review whether the note was sent so nobody double sends.

Include a per agent tab. For each agent, trend their average review score over time from the reviews stored in the app, broken out by the four rubric areas so a lead can see whether the problem is greetings or resolutions, and show it next to their LiveChat numbers from Report: Chat Ranking and Report: Chat Ratings over the same period. The point is to answer whether last month's coaching actually moved the number, so put the app's review scores and LiveChat's own rating data on the same timeline, and show how many chats were reviewed behind each average so nobody reads too much into a single scored chat.

Two implementation notes. LiveChat reads are POST calls even for the list methods, so List Archives, Get Chat, List Threads and List Tags all go out as POST with a JSON body. The paged list methods return next_page_id, which the handler should pass back as page_id to fetch the following page, so make sure the queue can load a full week rather than stopping at the first page.

## How to customize

- Change the review window and the queue rules, for example pull the last two weeks instead of one, or float any chat over ten minutes to the top
- Rename the four scoring areas, adjust the scale, or set the passing score that decides which review tag gets applied
- Pick the tags written back to LiveChat, and choose whether accepting a review sends the coaching note to the agent in Slack automatically or only when the reviewer opts in

## FAQ

### Does the AI grade replace my reviewers?

No. The grade is a draft. It scores the conversation, shows the exact customer and agent lines it based each score on, and writes suggested coaching notes. Nothing counts until a human reviewer accepts it, and they can override any score or rewrite any note first.

### Can two people end up reviewing the same chat?

No. Each conversation is claimed by one reviewer and their scores are saved against their name, so a chat that has been picked up or already scored drops out of everyone else's queue.

### Does this change anything inside LiveChat?

Only what you ask it to. Reading and scoring happens inside the app. When a reviewer accepts a review, the app adds your chosen review tag to that conversation in LiveChat so the result is visible to anyone working in the archive, and it can also store the verdict on the conversation itself.

### Do my agents need a login to get their feedback?

No. Agents do not need to open the app. Coaching notes reach them as a private Slack message, and the per agent tab is there for whoever runs the reviews.

### Can we use our own scoring rubric?

Yes. It ships with greeting, diagnosis, tone and resolution because those cover most support reviews, but you can rename the areas, change the scale, and set your own passing score. The AI grader is told to use whatever rubric you land on, so its scores stay comparable to your reviewers'.

### How far back can we review?

As far back as your LiveChat history goes. The queue defaults to last week's finished conversations because that is the usual review rhythm, and you can widen the date range whenever you want to look at a longer stretch.

Use this prompt in General Input: https://www.generalinput.com/prompts/chat-quality-review-board-for-livechat-support-leads