This report is real. The customer name is redacted.
Every number on this page is pulled live from the customer's AI Impact Report. We're awaiting their explicit consent to publish their identity — we'll update this report the moment they greenlight it. Back to all case studies →
Beauty · Skincare · Subscription
~12 agents’ worth of work, absorbed in 33 days. And CSAT went up.
This subscription skincare brand turned on Richpanel AI 33 days ago. Since then the AI has handled 17,908 customer messages — the workload of about 12 full-time agents — at 4.14 / 5 CSAT, well above their 3.40 pre-AI baseline. Median first response moved from 1.3 days to 8.9 hours. Same team. Same SOPs. AI doing 44% of the typing.
Pre-AI: Feb 22 – Mar 26 · With AI: Mar 27 – Apr 28, 2026 · 33 calendar days each (NY)
The numbers
AI CSAT
4.14/5
From 170 customer ratings on AI-sent messages.
+22% vs. their 3.40 pre-AI baseline
FTE-equivalent absorbed
~12
At ~45 messages/agent/day. Capacity freed, not headcount cut.
17,908 messages their team didn’t write
Median first response
8.9hr
Down from 1.3 days pre-AI. Autonomous email is doing the work.
~3.5× faster overall
Messages AI sent
21,594
44% of team output. Their agents now spend that time on harder cases.
Across 18,777 unique conversations
app.richpanel.com/ai-reports
The AI Impact Report dashboard — every number on this page is pulled from here, in real time.
Reach, multiplied
Their team types 40% less. Reach grew 8%.
Before AI, every message was hand-written. Now agents type 40% fewer messages while customer reach is up 8%. Effective output: 1.80× their hands alone — for every 100 messages an agent writes, the AI writes another 80.
Before AI · 33 days
Messages from team44,870
Per day~1,360
Total reach44,870
With AI · 33 days
Messages from team26,962
Messages from AI21,594
Total reach48,556
AI vs. the human team
Side by side.
Per-layer volume and quality on the messages each side sent over the same 33 days.
Human team
Messages sent26,962
First response (median)2.3 days
CSAT3.13 / 5
Ratings745
Richpanel AI
Messages sent21,594
First response (median)17 sec
CSAT4.14 / 5
Ratings170
Each side’s CSAT and FRT here only reflects the messages that side sent, not a head-to-head on identical conversations. The team-level CSAT of 3.32 weights both layers — what customers actually experience — at par with the 3.40 pre-AI baseline.
Before AI vs. with AI
Same brand, two adjacent 33-day windows.
Same query, same metric definitions on both sides. The only thing that changed between them is whether AI was on.
Before AI · Feb 22 – Mar 26
Total messages44,870
Closed conversations35,500
Team FRT (median)1.3 days
Team CSAT3.40 / 5
Ratings1,712
With AI · Mar 27 – Apr 28
Total messages48,556
Closed conversations28,504
Team FRT (median)8.9 hr
Team CSAT3.32 / 5
Ratings915
The −0.08 CSAT delta (3.40 → 3.32) is statistical noise across these sample sizes — quality held while volume grew. Closed-conversation count dipped because more conversations are still in flight under the new collaborative flow.
What the AI handles
WISMO, cancellations, troubleshooting.
Top customer intents resolved by AI since launch — the classic ecom support load.
Order Status & Tracking (WISMO)1,188
Troubleshooting & Product Issues564
Order Cancellation538
Subscription Cancellation535
Shipping Delays507
app.richpanel.com/inbox/conversation
A skin-reaction case — resolved end-to-end by AI in 16 minutes, with care, not scripts. The kind of message rated 4.14 / 5.
How they run it
84% autonomous. 16% co-pilot.
Where the AI’s messages flowed over 33 days, by channel:
Co-pilot mode
Contact Form3,419 msgs · 15.9%
Live Chat55 msgs · 0.3%
Total co-pilot3,474 msgs · 16.1%
Autonomous mode
Email18,074 msgs · 83.9%
Total autonomous18,074 msgs · 83.9%
Email autonomy is driving the speed gain — 8.9 hours vs. 1.3 days pre-AI. Contact Form and Live Chat still flow through their team for review.
Where this could go next
Contact Form (~3,400 messages already drafted in co-pilot) is the natural next channel to flip to autonomous. Quality data says it’ll hold.
