Before building Deskly AI, I went through support inboxes from 40 different stores doing between $10k and $500k a month. I wanted to know what store owners were actually spending their nights answering.

The breakdown was almost identical every single time:

1. "Where is my order?" — 31%

The customer has a tracking number. They just don't trust it, or the carrier hasn't scanned it yet. Nothing about this requires a human.

2. "Can I change or cancel my order?" — 17%

Almost always within 2 hours of purchase. If you answer in 20 minutes instead of 10 hours, you save the sale. If you answer the next morning, it already shipped and now it's a return.

3. "What's your refund policy?" — 12%

The policy is on the site. They didn't read it. They will never read it. They will always ask.

4. "Does this fit / is this compatible?" — 8%

This one is actually worth money. It's a pre-purchase question, and how fast you answer it directly decides whether they buy.

That's 68% of the volume, and only the last one genuinely needs judgment.

Here's what most owners get wrong: they hire a VA for $1,500-2,500 a month whose entire job becomes copy-pasting tracking links. Meanwhile question #4 — the one that actually converts — sits unanswered for six hours because the VA is asleep in a different timezone.

The fix isn't more headcount. It's this order of operations:

• Automate the 60% that's pure lookup (order status, tracking, policy)

• Route pre-purchase questions to a human immediately, day or night

• Track which questions keep coming back and fix the product page causing them

That third one is what nobody does. Every repeated question is a broken page somewhere. If 200 people ask about sizing, your size chart is the problem, not your support team.

That insight is basically why I built Deskly — not to replace support, but to stop owners drowning in lookups so they can answer the questions that actually make money.

If you run a store, go count your last 100 tickets and sort them into those four buckets. I'd bet money your split looks almost exactly like mine.