tradeit.gg · support operations

How the daily support digest is built

A plain-language explainer of the automated Intercom digest delivered to Slack each morning — no engineering knowledge required.

An automated summary of the previous 24 hours of support conversations — what happened, what needs attention, and how the team is performing — without anyone having to open Intercom.

Time window

What each digest covers

Each digest covers a rolling 24-hour window ending at approximately 06:00 UTC. So the Friday morning digest covers conversations from Thursday 06:00 UTC through Friday 06:00 UTC.

The exact start and end times are shown at the top of every digest.

How conversations are gathered

Two sets feed two different parts of the report

Opened in the window

Conversations a user started during the 24-hour window. These are analyzed for categories, tone, and emerging themes.

Replied to in the window

Conversations where an agent sent a reply during the window, regardless of when it opened. These count toward each operator's scorecard.

How tone is decided

An AI model reads the user's messages and judges their mood at the end of the exchange

Positive
Neutral
Negative
Abusive

What "Neg %" means

On the operator scorecard

The Neg % column shows the share of an operator's conversations that ended in a negative or abusive tone.

It is calculated only against conversations that received a tone rating — the ones opened and analyzed in the current window. Conversations marked "from older tickets" are excluded. That's why the percentage is out of the rated subset, not the operator's total count, and the tone numbers can be smaller than the total.

Categories & emerging themes

Known buckets, plus candidates for new ones

Known categories

Every analyzed conversation is sorted into a known support category (e.g. "Trade Failed" or "Deposit Issue").

Emerging themes

Conversations that don't clearly fit any existing category are grouped and surfaced under "Proposed new category" — candidates that may be worth formally adding.

The spike marker

When a category gets unusually busy

A category is marked with a flame (, shown as 🔥 in Slack) when today's volume is at least double its recent daily average and the average is at least 3 conversations per day. It signals an unusual surge worth investigating.

Each category row also carries an up/down arrow showing whether today is above or below its typical daily volume.

The AI agent (Tray)

Why a bot appears in the scorecard but never in "Top performers"

An automated bot called Tray handles first contact on most tickets. Because it processes a very high volume automatically, its numbers are listed in the operator scorecard for transparency — labeled (AI agent) — but excluded from "Top performers" recognition.

High volume from an automated agent is automation coverage, not a performance achievement.

Ticket links

Everything actionable is one click away

Every conversation reference in the digest is a clickable link that opens the conversation directly in Intercom — example tickets in "Needs attention", stuck-queue tickets, every abusive-tone conversation, recognition cases, and curiosity references. The digest never prints a raw conversation ID.

Where to go for more detail

Full evidence behind every number

The complete categorized evidence for each day is in the linked audit file at the bottom of every digest. The operator audit file (linked from the leadership digest) shows the complete per-operator breakdown, including the exact messages the model saw and why it scored each tone.


tradeit.gg support operations · generated from docs/how-the-digest-works.md