A sunlit plaster wall with a single reclaimed-wood shelf running across it, the shelf empty

Case studies

Two on the shelfAI customer support case studies, written without invented customers

About these two

We do not publish named customer case studies yet. Every business on this page is a composite of deployments in that sector, with the company and the people deliberately generic, and the before-and-after figures written as directions rather than measurements. When we have a named customer willing to be quoted with numbers we can stand behind, it will sit here with their name on it and not before.

2 of 2 volumes

Volume One

D2C health & hygiene

The inbox that never closed

A growing D2C brand taking hundreds of support requests a day across WhatsApp, email and Instagram, with a support team too small to answer them inside the hour.

What was happening

Order-status questions, return requests and product FAQs arrived faster than anyone could open them, and they arrived in the evening rather than the working day. Every unanswered message was a customer deciding for themselves whether the brand was reliable.

What was built

A WhatsApp agent grounded in the brand's own product catalogue, return policy and FAQ document. It answers order status, walks a customer through a return, sends the payment or tracking link mid-conversation, and routes anything it is unsure about to a human with the whole thread attached. Set-up work is measured in days, not quarters, because the knowledge base is the brand's existing documents.

First response

Hours

Minutes

Handled without a human

None

The routine majority

Cover

Office hours

Every hour

Time to go live

Under a week

Direction of travel for a deployment of this shape — an illustrative model, not a measured result from a named customer.

How the WhatsApp agent works

Volume Two

Real-estate discovery platform

The enquiry that arrived at 9pm

A property discovery and rental platform losing leads to the gap between when a buyer enquires and when an office opens.

What was happening

Buyers and tenants browse in the evening and enquire the moment something interests them. A reply the next morning arrives after they have already spoken to somebody else. The leads were not lost to a better product; they were lost to a ringing phone.

What was built

An agent across WhatsApp and voice that answers on the first ring at any hour, answers questions about a listing from the documents the platform uploads, captures what the buyer is actually looking for, sends the brochure, and books the site visit into the calendar — handing over to a person with the full transcript when the conversation needs one.

After-hours enquiries answered

A fraction

All of them

Pre-sales questions handled by the agent

None

The routine majority

Site visits

Booked next morning

Booked in the conversation

Lead response

Overnight

Under a minute

Direction of travel for a deployment of this shape — an illustrative model, not a measured result from a named customer.

What the agent does for real estate

What carries across both

The knowledge base is your own documents: Nothing is hand-written into a bot script. Upload the return policy, the price list, the brochure, and the agent answers from those — so the answer changes when the document does.
The handover carries the context: When a conversation needs a person, it transfers with the transcript and the caller's intent, so nobody is asked to repeat themselves to a second human.
One record per customer: Voice and WhatsApp for the same person land on the same record, which is what makes the second conversation better than the first.
The language is theirs, not yours: All 22 scheduled languages of India, including the Hindi-English mix people actually speak on the phone.

Put your own on the shelf

Book a demo and we will walk your own workflow through an agent, on your documents, in the channel your customers already use.

Book a demo Start free instead