Case Study — AI Agent Development

Turning WhatsApp from a support
backlog into a sales channel.

Client name withheld under NDA. This is a real engagement — the client name "Thornby & Finch" is a placeholder, and some identifying details have been generalized, under the confidentiality terms of our agreement with the actual client. Outcome figures below reflect typical ranges reported for WhatsApp-based AI support and sales automation in e-commerce, rounded and generalized to protect the client's specific metrics — not precise measurements of this engagement.
Client: Thornby & Finch (name changed) Industry: E-commerce / DTC retail Timeline: ~6 weeks Related: AI Agent Development
The problem

Customers were already on WhatsApp. The team just couldn't keep up.

Thornby & Finch's customers didn't email or call — they messaged the brand's WhatsApp Business number, the same way they'd message a friend. "Where's my order," "does this come in a bigger size," "is this back in stock" — a mix of support questions and pre-purchase sales questions, all landing in the same shared inbox.

During business hours, a small team answered manually. Outside those hours — evenings, weekends, and almost everything from customers in other time zones — messages just queued up. Support questions sat unanswered for hours. Sales questions were worse: a shopper asking "will this fit a king bed" at 9pm got an answer the next morning, by which point many had already bought from someone else or lost interest entirely.

What it was built on.

The setup as we found it:

WhatsApp Business number, manually staffed Shared inbox tool, no automation Order and inventory data in a separate backend, not connected to chat No triage between quick-answer and judgment-call messages No coverage outside business hours
The approach

Let the agent answer what it can verify. Hand off the rest with full context.

The goal wasn't to automate the whole inbox — it was to stop routine questions from competing with the ones that actually needed a person:

The backlog stopped building up overnight.

Figures are generalized industry ranges for this category of deployment, not exact client metrics, per our NDA.

Before
Messages outside business hours queued for hours; sales questions often answered too late to matter
After
Most routine order-status, product, and sizing questions resolved immediately, any time of day
Ongoing
Support team now focused on escalations and judgment calls, not repetitive lookups
Related

If this looks familiar.

This is the same pattern we cover in why AI customer support agents get escalated back to humans, and the kind of deployment our AI agent development work is built to get right — answer what can be verified, hand off what can't.

Get started

Tell us what
you're trying to build.

Book a 30-minute call — we'll tell you honestly whether an agent on WhatsApp (or wherever your customers already are) makes sense for you.

Book a free 30-min call → More case studies →