Skip to content
BlendDesk

Industry · Ecommerce

Customer operations for stores that run on messaging.

Order questions, returns, delivery, and the four jobs underneath them. One thread per customer across WhatsApp, Instagram, Messenger, email and web chat, with the order record attached and an agent your team can rehearse, approve and audit.

A store inbox in which three conversations arrive and settle differently: an order question answered from the order record, a product question answered from approved knowledge, and a return handed to a named agent because it is over the refund limit.

Illustrative demo with fictional data.

A morning in the inbox

Three conversations, three different endings.

The useful question is not how many conversations an agent closes. It is which ones should ever have reached an agent.

  1. Order status, answered from the order

    A WhatsApp question about an order gets the courier, the tracking number and the delivery window, cited to the order record Scout read.

  2. A product question, answered from knowledge

    An Instagram DM about a finish and stock level is answered from the catalogue and knowledge you approved, with the sources on the reply.

  3. A return, handed to a person

    A refund is over the limit you set, so the thread is assigned with the full history, the order and Scout's draft attached. The customer is told a person is taking over.

Answer → Resolve → Act → Improve

The lifecycle of one order question.

The same four stages run behind every job on this site. On a store, they look like this.

Answer
Five channels land in one thread, attached to a contact with its order history. Nobody asks which app the customer used, and nobody answers the same question twice because they could not see the first answer.
Resolve
Scout answers from the knowledge and catalogue you approved, cites what it used, and hands over when the case is outside its scenario or over a limit.
Act
Confirmations, order lookups, address changes, refunds and reminders execute against your store — with approvals on anything that moves money.
Improve
Replay real past conversations as an evaluation, read the score, read the failures, change the scenario, and ship the change having already seen what it would have said.

Message economics

Your biggest controllable cost is the template category.

Most stores discover this on an invoice. BlendDesk models the 24-hour window and the template category directly, and shows you the category before the message leaves.

Open the WhatsApp cost calculator
Utility template (Egypt)
$0.0036
Order confirmations and shipping updates — messages about an order the customer already placed.
Marketing template (Egypt)
$0.0644
Promotions and cart reminders. About eighteen times the utility rate for identical text.
Inside the 24-hour window
Service message
A message from the customer opens a 24-hour window. Meta began charging for service messages on 1 October 2026, with the first 1,000 per business phone number per month free.

Source: Meta WhatsApp per-message rate card, Egypt, retrieved 2 November 2026; category, window and service-message rules from Meta developer documentation.

Meta bills these fees, not us, and we never mark them up. BlendDesk charges for seats and AI resolutions.

What we will not tell you

The numbers we refuse to publish.

No return-rate claim
You will find vendors quoting an exact percentage by which confirmation messaging cuts returned shipments. We found no source for any of them that was not a vendor selling the feature, so we publish none.
No resolution rate
Everyone in this category advertises one. None of them let you verify it. Instead you get a playground, evaluations against your own past conversations, and a trace per run.
No customer logos
BlendDesk is new and has no customers to name. When that changes we will name them with their permission and their own numbers, not a stock logo wall.
No markup on Meta's fees
Meta bills message fees directly. We charge for seats and AI resolutions, and we show Meta's published rates so you can predict the other half of the bill.

Questions buyers actually ask.

Do you integrate with my store?

Order data reaches BlendDesk through a native integration or through a custom HTTPS tool pointed at your own endpoint. Either way the call appears in the run trace with its response, so you can see exactly what the agent read.

Will Scout ever issue a refund on its own?

Only if you decide it should, and only under the limits you set. By default a refund is an approval: Scout proposes it, a person signs it, and the approval is in the trace.

We already use a shared inbox. Why move?

Most shared inboxes are email-shaped, with WhatsApp bolted on as another channel. The 24-hour window, template categories and order actions are not modelled, so your team manages them by memory.

How much of this can run without a person?

As much as you are willing to rehearse and approve. The honest answer for a new store is: start with order status and confirmation, read the traces for two weeks, then widen the scenario.

Does it handle Arabic?

Yes, including Arabizi and mixed Arabic-English threads, with a right-to-left interface. It is engineering we did rather than a translation layer, and it is a capability rather than our position.

Start with the job that is costing you the most.

Book a demo and we will run order confirmation or order status against a test store, with your templates and your rules.