Atlas Goods: from 200 Zendesk macros to zero

Atlas Goods sells outdoor equipment to 1.4M customers across three brands. Their 32-person support team was processing 4,800 tickets a week, and 60% of those tickets started life as a Zendesk macro. When we started working with them in November, their CSAT was 78. Five months in it's 92. Here's what changed.

The macro problem

Macros are insidious. They start as time-savers — "answer the shipping question once and reuse it." Three years in, you have 200+ macros, half of them outdated, and agents spend more time picking the right one than writing the reply would have taken. Worse, customers can smell a macro at thirty paces. Atlas Goods' CSAT comments mentioned "copy-paste" or "template" 312 times in the quarter before we engaged.

Week 1–2: shadow mode

We turned on QuickPly in shadow mode — drafts generated, but invisible to agents. We compared what the AI would have sent against what the macros said. The drafts won 71% of the time on length-adjusted readability and 84% on tone match. That was the green light.

Week 3–6: kill the macros

1. Inventoried all 217 macros. Categorized by use-frequency.
2. Top 20 macros covered 73% of usage. Trained the tone fingerprint on the actual sent replies that followed those macros, not the macros themselves.
3. Deleted 197 macros outright. Kept 20 as fallbacks for after-hours auto-acknowledgement.
4. Rolled out QuickPly to the whole team on a Tuesday. Tuesdays are easier than Mondays.

The change wasn't smooth. Three agents quit in the first month — but two of them had been with the company under six months, and the third said she felt the AI was making her job too easy. The remaining 29 are still there, and per-agent throughput is up 41%.

The CSAT lift

CSAT went from 78 → 92 over five months. We can't attribute that entirely to QuickPly — they also redid their returns flow in February — but the comment analysis is striking. "Copy-paste" mentions dropped from 312 a quarter to 41. "Personal" and "thoughtful" mentions roughly doubled.

"I don't think we'd go back. The macros made us feel productive. QuickPly actually makes us productive."

What we'd do differently

We underestimated change management. Killing 200 macros is a cultural event, not a technical one. If we ran the playbook again, we'd start with a smaller pilot team — maybe 6–8 agents — for a full month before rolling to the whole org. The data would have been just as convincing, and we'd have lost zero people instead of three.