Customer story
Replies in minutes, launches measured in orders.
Alma Concept, a skincare brand from Uruguay, runs its customer conversations and its WhatsApp launches on Kanesh.
Results
- Attributed
26.9%
of the September launch recipients clicked and ordered within 7 days
819 buyers out of 3,046 recipients, 12–18 September 2026. The August launch, sent to a Klaviyo segment built for it: 15.2% (908 out of 5,955).
- Estimated
24.4%
of paid Shopify orders came within 48 hours of a conversation
854 out of 3,504 orders, 1–29 September 2026, matched with a 48-hour window. AI alone 307, AI and a person 290, a person alone 257.
- Measured
2.2 min
median first reply on WhatsApp
2,562 conversations started by customers, 1–27 September 2026. 96.3% got a reply within 24 hours; the AI replied first in 1,442.
The challenge
Alma sells a skincare routine, and much of that selling happens in chat. Launch days bring a wave: on 20 April 2026, 5,435 conversations came in, against 353 on a normal day.
With the previous tool, the median first reply from a person was 46 minutes on a normal day, and on 1 June 2026 it reached 23.5 hours. Instagram, one of the two main channels, had no way to link a chat to an order.
That baseline counts the first reply from a person in the previous tool, so we don't compare it with the figures below.
What we did
In Kanesh
- One shared inbox for WhatsApp, Instagram, Messenger and email, where the team and the AI work the same conversations.
- An AI agent that answers from Alma's policies, catalogue and verified Shopify orders, and hands the conversation to the team with the reason when it shouldn't continue.
- Automatic follow-ups when a customer goes quiet.
- WhatsApp launch messages to lists built in Klaviyo for each launch, with tracked links, so every order after a click is counted.
- Welcome flows triggered from Alma's Klaviyo lists.
Nexau services
- Moved the conversation history over from the previous WhatsApp tool.
- Set up the knowledge and the handoff rules, then read 680 handoffs one by one (8–16 September) to remove the unnecessary ones.
- A separate Nexau project, not part of Kanesh: Alma's commercial-intelligence dashboard and operations automations.
How we measured
- Launches: WhatsApp template messages with tracked links. A buyer counts when they clicked the message's link and then placed a paid order (refunds excluded) within 7 days of the send.
- Orders after a conversation (estimated): an order counts when the same customer, matched by contact, phone or email, had a conversation in the 48 hours before paying. The split shows who wrote in that conversation.
- First reply: a conversation starts when the customer writes after 24 hours of silence. The first reply is the first message from the AI or a person within 24 hours. Templates and automatic away messages don't count.
- Source: Kanesh production data, queried read-only on 30 September 2026. No revenue figures are published.
Limits
- Attributed and estimated are not incremental. There was no hold-out group, so some of these customers would have bought anyway.
- The September audience had signed up to hear about the launch. Messages to people who didn't ask convert far less, so don't read these rates as typical broadcast results.
- The 48-hour match needs a phone number or an email. Instagram chats often have neither, so orders that started there are under-counted.
- The two launches went to different lists of different sizes, so the gap between them isn't a controlled comparison.
In their words
See what your conversations are missing.
A free, written review of up to 180 days of your WhatsApp history, within 48 hours.