AI customer support in e-commerce: 2% to 94% in four months
Updated September 2026
A case study in numbers: one e-commerce brand put Sky, CallBolt's AI agent, on its support tickets and phone line, and this is what changed.
What did the deployment look like?
The client started Sky on one department's ticket queue: order status, cancellations, billing questions and booking changes. As policies and knowledge tightened, coverage grew week over week, and a second department (online sales) came on and reached 79% resolved by Sky in its first month.
How did resolution share move month by month?
| Period | Share of tickets resolved by Sky |
|---|---|
| April 2026 | 2% |
| May 2026 | 61% |
| June 2026 | 71% |
| July 2026 | 77% |
| Latest 24-hour window measured | 94% |
The weekly view inside the ramp told the same story: the resolution rate rose every week for four weeks, from 58% to 78%.
What did it return to the team?
| Measure | Number |
|---|---|
| Tickets received in 90 days, one department | 39,718, of which 99.2% closed overall |
| Tickets fully auto-resolved by Sky | 28,252 |
| Human hours returned to the team | about 5,268 |
| Closures touched by Sky in the second department | 97% |
| Voicemails transcribed and worked without a human listening first | 326 of 327 |
What should you not read into these numbers?
They are one deployment, not a promise. The ramp started at 2% because the policies started narrow, and it climbed because the client kept tightening them. A deployment that skips that work does not get the curve.
The same agent runs phones and tickets together; the flow is on the AI customer support agent page, and the plan shape is on CallBolt pricing.
Frequently asked questions
Did Sky replace the support team?
No. Sky took the routine ticket volume and the phone line, and the team kept the cases that need a person. 28,252 tickets were fully auto-resolved over the period, which returned about 5,268 human hours to that team.
How long did the ramp take?
Four months from 2% to 94%: 2% of tickets resolved by Sky in April, 61% in May, 71% in June, 77% in July, and 94% in the latest 24-hour window measured. Coverage grew as the client's policies and knowledge tightened.
What kinds of tickets did Sky resolve?
The everyday e-commerce volume: order status, cancellations, billing questions and booking changes, plus voicemails transcribed and worked like tickets (326 of 327 handled without a human listening first).
Can we get numbers like this for our store?
The shape of the ramp is the honest expectation: start small, tighten policies, earn volume. Your numbers will be yours, scoped with the team; a quote usually comes back within one business day.
Which channels did Sky cover in this deployment?
The support ticket queue and the phone line, plus voicemails, which were transcribed and worked like tickets: 326 of 327 handled without a human listening first.
What did the second department reach?
The online-sales department reached 79% of tickets resolved by Sky in its first month, and Sky handled 97% of that department's closures.
Want this curve on your own queue?
Tell the team your ticket and call volumes, and we will scope what the first month looks like for you, usually within one business day.
We usually reply within one business day.