# Weekly coaching reviews for every Dixa support agent

> Every Monday we review a sample of last week's closed conversations, DM each agent private feedback, and log scores to a shared scorecard.

- Workflow type: agent
- Services: Dixa, Google Sheets, Slack Bot
- Categories: Customer Support, Operations
- Published: 2026-08-14

## What it does

- Picks a handful of last week's finished conversations for each agent and reads them start to finish, including the satisfaction rating the customer left.
- Grades each one on the things that matter: tone, whether the question actually got answered, how clear the resolution was, whether next steps were set, and whether the reply matched how urgent the issue was.
- Sends every agent a private Slack message with two things they did well and one thing to change this week, quoting their own strongest and weakest reply.
- Logs a row per conversation to a Google Sheet so quality trends build up week over week, then posts a team roll-up to your support leads channel with no individual scores named.

## What you'll need

- A Dixa account with administrator access, so the review can read closed conversations, messages, and satisfaction ratings
- A Slack workspace where the bot can send direct messages and post in your support leads channel
- A Google Sheet to hold the scorecard, with a tab ready for the weekly rows
- Agent email addresses that match between Dixa and Slack, so each review reaches the right person

## Prompt

Every Monday at 8am, run a support quality review that coaches my Dixa agents instead of just reporting numbers about them.

Start by using the Dixa "Search conversations with filters" operation to pull every conversation that was closed in the last 7 days. Then use the Dixa "List teams" operation, and "List team members" for each team, to build a lookup of who handled what, including each agent's name and email address.

Group the closed conversations by the agent who handled them. Skip any agent with fewer than three closed conversations that week rather than scoring them on a thin sample. For every remaining agent, take a fixed sample of their conversations, five by default, spread across the week rather than all from a single day.

For each sampled conversation, read the full thread with the Dixa "List messages" operation and pull the customer's score with the Dixa "List ratings for a conversation" operation. Ratings are fetched one conversation at a time, so pace the requests to stay inside Dixa's limit of 10 requests per second. The per-agent sample cap is what keeps the total volume manageable, so do not remove it.

Score every sampled conversation against a plain quality rubric: greeting and tone, whether the customer's actual question was answered, resolution clarity, whether next steps were set, and whether the reply matched the urgency of the issue. Score each of the five on a 1 to 5 scale and write a one-line note explaining the score. For each agent, identify their single strongest and single weakest reply across the sample and capture each one as a verbatim quote, so the feedback is anchored in their own words rather than in a generalization.

Log one row per reviewed conversation to my Google Sheets QA scorecard using the "Append Values" operation. Each row should carry the review date, the agent name, the conversation id, the five rubric scores, the satisfaction rating, and the one-line note. Always append rather than overwrite, so quality trends build up week over week and I can compare an agent against their own past weeks.

Then coach privately. For each reviewed agent, take their Dixa email address and use the Slack Bot "Look Up User by Email" operation to find their Slack account, "Open a Conversation" to open a direct message with them, and "Send a Message" to deliver their personal review. Each message should give two specific things they did well and one concrete thing to change this week, including the verbatim quotes from their strongest and weakest reply. Keep the tone supportive and specific, never generic praise, and tie every comment to something they actually wrote.

Finally, post a short team-level roll-up to my support leads channel using the Slack Bot "Send a Message" operation. Cover the common misses across the whole sample and what the team should focus on this week. Do not name individual agents or their scores in the public channel. Individual feedback stays in direct messages, and the channel post stays at the level of themes and patterns.

## How to customize

- Change the day and time it runs, or move it to a fortnightly cadence if weekly feels like too much
- Raise or lower how many conversations get reviewed per agent (five by default) and the minimum someone needs before they are reviewed at all (three)
- Swap the qualities being graded, or point the team roll-up at a different channel

## FAQ

### Will agents be able to see each other's scores?

No. Individual feedback only ever goes to that agent's own direct message. The message posted in your leads channel covers common themes across the whole team, with no names and no individual scores attached.

### What happens to someone who only handled one or two conversations last week?

They get skipped that week. Judging someone on a single conversation produces noise rather than coaching, so anyone below three closed conversations is left out until they have enough of a sample to be fair.

### Do I need a paid Dixa plan for this?

You need administrator access in Dixa, since that is what allows the review to read your conversation history and satisfaction ratings. Any plan where you have admin rights will work.

### Can it review every conversation instead of just a sample?

You can raise the number reviewed per agent, but a sample is deliberate. It keeps the review quick, keeps it inside Dixa's usage limits, and five well-read conversations produce better coaching than a hundred skimmed ones.

### What if an agent's Slack and Dixa email addresses are different?

People are matched by email, so if the two differ their Slack account will not be found and their review will not send. Keeping the addresses consistent across both tools is the simplest fix.

Use this prompt in General Input: https://www.generalinput.com/prompts/weekly-coaching-reviews-for-every-dixa-support-agent