How to Surface Daily and Weekly Trending Support Topics (Automatically)

Surfacing trending support topics is the discipline of catching that pattern early: knowing which issues are growing, which are concentrated, and which are accelerating, before they become a flood. Done well, it splits into two jobs. The first is real-time spike detection — a fast alert when one topic suddenly spikes. The second is the daily and weekly read — a steady digest of what your customers are actually contacting you about, so you can fix root causes instead of mopping up symptoms.

This post shows how to do both with Macha. To be clear about what Macha is: it's an AI agent layer that sits on top of the helpdesk you already run — Zendesk, Freshdesk, Gorgias, Front — not a replacement for it. Your tickets stay where they are. Macha adds agents that read those tickets on a schedule, classify them by topic, and push the result to the channel your team actually watches: Slack.

The two patterns that matter

Good trend monitoring isn't one feature; it's two cadences working together.

The mistake most teams make is treating volume as the trigger. As Cobbai's anomaly-detection guide puts it, you shouldn't alert on "volume is high" — you should alert on "volume is high and concentrated": one topic growing fast, one region spiking, one channel misbehaving. A good alert answers three questions immediately — what changed, where it's concentrated, and whether it's accelerating. Keep that framing in mind; it's the difference between a useful alert and an alarm everyone learns to ignore.

Pattern 1: real-time spike alerts to Slack

This is the Notify team of trending support topics use case, and it's the one that earns its keep on a bad day.

The shape is simple: a Macha agent wakes up on a schedule, searches your recent tickets, compares the topic mix against what's normal, and — only if something is genuinely spiking — posts an alert to Slack. Two connectors do the work: Zendesk Search Tickets to read the queue, and Slack Send Message to raise the flag.

Step 1 — Put the agent on a clock

Macha agents can run on a scheduled trigger: cron-based execution on a recurring interval. You don't have to babysit it. A few guardrails are baked in so a clock-driven agent can't run away with itself:

• Minimum interval is 1 hour, with a cap of 24 runs per day and an org-wide concurrency limit, so scheduled agents stay predictable.
• If a scheduled trigger fails 5 times in a row, it auto-disables and emails you — a runaway job stops itself instead of quietly burning credits.

For spike detection, hourly is the sweet spot. You catch an emerging incident within the hour without paying for a check every five minutes that almost always finds nothing.

Step 2 — Tell the agent what "spiking" means

The intelligence lives in the agent's instructions. You give it the rule in plain language — for example: search the last two hours of tickets, group them by topic, and if any single topic is running 3× its typical hourly volume, raise an alert. Because the agent reads the actual ticket text (subject, description, the comment thread), it isn't limited to tags that were applied inconsistently months ago. It clusters the real language customers are using — "can't log in", "502 error", "card declined" — into a topic on the fly.

That's also how you keep false positives down. Instead of "12 tickets is a lot," the instruction is "12 tickets about one thing, when that thing usually generates 3 — and they all mention a 502 error." High and concentrated.

Step 3 — Post an alert worth reading

When the threshold trips, the agent posts to Slack. A good alert doesn't just say "topic up" — it carries the context that lets someone act in ten seconds:

🔥 Trending Topic Alert "Login issues" is spiking — 12 tickets in the last 2 hours (normal: ~3). Common thread: customers reporting a "502 error" on the login page. This may indicate a service issue. Consider checking with engineering.

That's the what changed / where concentrated / accelerating test passed in four lines. The same Send Message tool that powers Macha's other Slack alerts — VIP-ticket pings, escalations — delivers it into whichever channel your team lives in.

Pattern 2: the daily and weekly topic digest

Spike alerts are for emergencies. The daily and weekly digest is for steering — the calm, recurring read that tells you what your support load actually consists of so you can fix the upstream causes.

The mechanism is the same scheduled agent, set to a slower beat. Macha ships this as the Daily support metrics summary in Slack use case: every morning at 9am, an agent pulls the previous day's ticket data from Zendesk and posts a tidy summary — created, solved, response times, CSAT, and the top topic — to a Slack channel. Set the same trigger to "Monday at 9am" and it's a weekly review instead.

📊 Daily Support Summary — March 4 - Tickets created: 47 · solved: 52 - First response (avg): 12 min · resolution (avg): 2.4 hrs - CSAT: 94% - Top topic: Order tracking (18 tickets) ✅ Cleared backlog — 5 more solved than created

The "Top topic" line is the trend signal. Watch it across a week and the story writes itself: if "Order tracking" leads three days running, that's not noise, that's a process to fix. Leadership gets the same view without anyone building a spreadsheet — which is the entire point. The weekly tracking metrics most guides recommend — ticket volume, backlog, first response time, top contact drivers — all fit naturally into this digest.

When you need real depth: run a Study

A Slack digest is great for the daily heartbeat. But when you want to quantify a trend rigorously — "what share of last month's tickets were really about billing, broken out by sub-topic?" — that's a job for Studies, Macha's batch-analysis tool.

A Study points an AI model at a batch of tickets (scoped by a Zendesk search and date range) and fills in a typed grid: one row per ticket, one column per question. For trend work, your columns are the topic taxonomy — a single-choice Category, a free-text Sub-topic, maybe a 1–5 Urgency. The AI reads every ticket and classifies it consistently, which is exactly what inconsistent human tagging never delivered.

The payoff is the Report view: it aggregates each column into a chart. Choice columns become ranked bar charts of your topics with counts and percentages; number columns become histograms. Now "trending topics" stops being a vibe and becomes a chart you can click into.

Run the same Study over two date ranges — last week vs. the week before — and the movement between the two breakdowns is your trend. That's the rigorous version of "what's trending," and it's the one you bring to a product or engineering review. (We walk through a full worked Study, with real numbers, in our Studies deep-dive.)

A practical cadence to copy

Here's a setup that covers a real team without over-engineering:

1. Hourly spike agent — scans the last 1–2 hours, alerts Slack only on a concentrated spike. Your early-warning system.
2. Daily 9am digest agent — yesterday's volume, response times, and top topic into a team channel. Your heartbeat.
3. Weekly Monday digest — the same digest aggregated over seven days, for the wider team and leadership.
4. Monthly Study — a full topic breakdown with charts for root-cause and roadmap conversations.

Patterns 1–3 are the same building block — a scheduled agent with Zendesk + Slack — on three different clocks. Pattern 4 is Studies. Together they give you the fast signal and the slow, defensible read.

Watch-outs and when this isn't the right tool

No tool is free of trade-offs. A few honest ones:

• Scheduled triggers are a paid-plan feature with hard limits. They're available on Professional (3 scheduled triggers) and Enterprise (20), with the 1-hour minimum interval. If you genuinely need sub-minute detection of a hard outage, a dedicated APM/status tool watching your infrastructure will beat ticket-volume analysis every time — tickets are a lagging signal of an outage. Use trending-topic alerts for things your monitoring can't see: confusing UX, a bad email send, a policy change customers hate.
• Every scheduled run uses credits. Each agent run consumes credits (per AI action — the cost depends on the model you pick, with a low-cost default). Hourly checks are cheap; minute-by-minute checks would not be, which is part of why the 1-hour floor exists. Match the cadence to the value.
• Thresholds need tuning. "3× normal" is a starting point, not gospel. Build your baseline with day-of-week and time-of-day in mind — a Monday-morning spike isn't an incident, it's Monday. Expect to adjust the rule in the first week so the channel stays trustworthy.
• Studies' Report view is Professional and Enterprise only. On Trial and Starter you'll see an upgrade card. The batch grid is the heavy lift; the charts are the Pro/Enterprise layer.

If your volume is tiny — a few dozen tickets a week — you may not need automation at all; you can read your queue. The value of this kicks in once the volume is high enough that nobody can hold the whole shape of it in their head.

How it fits with the rest of Macha

Trend monitoring is the "watch" half of a loop. The "act" half is where Macha agents resolve, triage, and route tickets directly on Zendesk and other helpdesks. The same Slack connector that delivers your alerts can also relay urgent tickets, answer FAQs, and route questions to the right channel. So the trending-topic alert that tells you "login issues are spiking" can sit right next to the agents that are already first-responding to those tickets — you see the trend and the queue is being worked at the same time.

FAQ

What counts as a "trending" support topic? A topic whose volume is rising fast and is concentrated — one issue, region, or channel behaving differently than the baseline. Raw high volume alone isn't a trend; concentration and acceleration are what make it worth an alert.

How quickly can Macha catch a spike? Scheduled agents run on a minimum 1-hour interval, so hourly is the fastest practical cadence for trend detection. For true real-time infrastructure outages, pair this with a dedicated monitoring tool — tickets lag the actual incident.

Does this work on Freshdesk or Gorgias, not just Zendesk? The pattern is helpdesk-agnostic. Macha connects to Zendesk, Freshdesk, Gorgias, and Front; the agent searches whichever helpdesk you've connected and posts to Slack the same way.

Do I need Studies for this? No — the Slack digest covers daily and weekly trending topics on its own. Studies is for when you want a rigorous, chartable breakdown across thousands of tickets, which is overkill for a daily ping but ideal for monthly root-cause analysis.

How much does it cost to run? Scheduled agents and Studies run on credits (priced per AI action, by model — a low-cost default keeps routine checks inexpensive). Scheduled triggers are a Professional/Enterprise feature. See the pricing page for current plans.

Get started

If you want to stop discovering issues from the size of your backlog and start catching them from the shape of it, this is a half-hour setup. Start a 7-day free trial, no credit card required, connect your helpdesk and Slack, build a scheduled agent with a topic rule, and let it watch the queue for you. The docs walk through scheduled triggers and Slack alerts step by step.

Written by Abbas (Customer Support & AI, Macha) · Reviewed by Ankeet Guha (Co-founder & CTO) · Published 2026-06-24 · Last updated 2026-06-24.