Should the first agent be customer-facing?

Usually not. Internal desks forgive a rough edge and tell you the truth. Customer surfaces go live after the write path and the escalation are proven.

Do you build a custom chat UI?

Only when the work needs a surface Slack and Salesforce cannot give. A new front end is a product, and we will say so before you fund one.

Is voice ready for regulated work?

For qualification, capture, and confirmation, yes, with a human on anything that commits money or identity. We will name the hand-off in the design.

Architecture

The agent has to live where the work already happens.

The first week is in the org, reading what already exists. Slack for the internal desk, Experience Cloud for the customer, voice for after hours. A surface nobody opens is not a channel. The honest no: if the work is a software factory with a named delivery date, we are the wrong partner.

Send the brief

What it is

Surfaces, not screens.

The experience plane is every place a person can start, watch, or stop a piece of digital labor. It has to acknowledge fast, show what the agent is about to do, and make the stop button obvious.

Why it matters

A good agent on a bad surface is shelfware.

Four ways this fails in practice. Each one is cheaper to fix before you build than after your team has decided the agent is useless.

In an org

A regional service organisation.

Three surfaces, one case record. This is the shape we usually implement first.

01
Slack for the desk
The team that already lives in Slack gets the agent as a command and an app home, not a new tool to learn.
02
Experience Cloud for the customer
Status, documents, and the next step behind a login, so the desk stops answering the same call.
03
Voice after hours
The call is answered, qualified, and written back. A person reviews anything the model should not sign.
04
One record
All three surfaces read and write the same case. Otherwise you have three inboxes and a reconciliation job.

Voice and chat are surfaces on the same case. The call is answered by a voice agent, the transcript and outcome write back through the governed path, and the desk sees one record in Slack or the console.

Mermaid source

flowchart TD
  A["After-hours call"] --> B["Voice agent: qualify, capture, confirm"]
  B --> C["Gateway: identity and policy"]
  C --> D["Case or lead in Salesforce"]
  D --> E["Desk in Slack or the console"]
  E --> F["Customer portal: status without a phone call"]

How Mindcat helps

Where we would start in your org.

One surface, one desk, one process. Not three channels in a quarter.

Questions

What we actually say.

Should the first agent be customer-facing?: Usually not. Internal desks forgive a rough edge and tell you the truth. Customer surfaces go live after the write path and the escalation are proven.
Do you build a custom chat UI?: Only when the work needs a surface Slack and Salesforce cannot give. A new front end is a product, and we will say so before you fund one.
Is voice ready for regulated work?: For qualification, capture, and confirmation, yes, with a human on anything that commits money or identity. We will name the hand-off in the design.

What to read next.

The brief

Name the surface people already use.

We will tell you whether the agent belongs there or somewhere quieter.

The agent has to live where the work already happens.

flowchart TD A["After-hours call"] --> B["Voice agent: qualify, capture, confirm"] B --> C["Gateway: identity and policy"] C --> D["Case or lead in Salesforce"] D --> E["Desk in Slack or the console"] E --> F["Customer portal: status without a phone call"]