One brain, many doors: what a desk is
The four parts of a desk — knowledge, doors, a pipeline and an exit — and why the exit is a part rather than a failure.
Most people building their first support agent build a chatbot: something that answers a question and then forgets it happened. It is genuinely useful for about a week, and then somebody asks why the CRM has no record of the forty conversations it had.
This course builds a desk instead. The difference is not intelligence — it is that a desk has a memory, a paper trail, and a way out.
The one idea. A chatbot is a conversation. A desk is an operation. Everything in this course hangs off the four parts below, and the capstone rebuilds all of them from scratch.
The four parts
Read it left to right and you have the whole course. People arrive through different doors. They meet the same agent with the same rules. It answers from what it knows rather than from memory. It leaves something behind. And when it cannot support an answer, it stops and says so.
| The part | What it is in Expedify | Why it exists |
|---|---|---|
| Knowledge | A Knowledge Base the agent searches | So answers come from documents you control, not from whatever the model absorbed in training. |
| Doors | Channels — web chat, WhatsApp, voice, email | A customer picks the door. They should not get a different quality of answer for picking WhatsApp. |
| A pipeline | CRM records — contacts, deals, notes, tasks | An answered question that changes nothing is a cost. The conversation has to move something. |
| An exit | A handoff: a task, an assignment, a named human | The most important behaviour your desk has. Lesson 4 is entirely about it. |
Knowledge
- What it is in Expedify
- A Knowledge Base the agent searches
- Why it exists
- So answers come from documents you control, not from whatever the model absorbed in training.
Doors
- What it is in Expedify
- Channels — web chat, WhatsApp, voice, email
- Why it exists
- A customer picks the door. They should not get a different quality of answer for picking WhatsApp.
A pipeline
- What it is in Expedify
- CRM records — contacts, deals, notes, tasks
- Why it exists
- An answered question that changes nothing is a cost. The conversation has to move something.
An exit
- What it is in Expedify
- A handoff: a task, an assignment, a named human
- Why it exists
- The most important behaviour your desk has. Lesson 4 is entirely about it.
Why the exit is a part, and not a failure
It is tempting to treat handoff as what happens when the agent breaks. It is the opposite. A desk that never hands off is not more capable — it is one that has been given no way to admit a limit, so it fills the gap with something plausible.
You will see this concretely in the very next lesson. An agent is asked a price it does not have; the helpful answer and the correct answer point in opposite directions, and only a designed exit gets you the correct one.
What this course is not. It is not about making an agent cleverer. Every failure you will meet here is a failure of boundaries — what it may say, what it may touch, when it must stop. The model is rarely the problem.
What you will have built by the end
- An agent that answers enquiries from your own catalogue and policies, and refuses to invent the rest.
- The same agent reachable on chat, WhatsApp and a phone call — one brain, not three.
- A pipeline that moves: deals change stage, notes get written, nothing gets recorded four times.
- An onboarding sequence that watches for absence, not just for events.
- A handoff a human can actually pick up, with the context already attached.
Next: the enquiry agent — and the single most useful habit in this whole course, which is teaching it that a fee is not prose.
Related lessons
Base rates — what a piece of evidence is actually worth
A face-recognition system that is 99.9% accurate and almost entirely wrong, and a number that sent an innocent woman to prison. Both are the same arithmetic, and it is the arithmetic that decides what any piece of evidence is worth.
ReadConfirmation and survivorship — what you never looked for
Two questions about evidence you did not go looking for. One is a rule you have to discover, and one is a pattern in five famous people — and in both, the thing that would have told you the truth is the thing nobody checks.
ReadLoss aversion, sunk cost and regression — what it costs you
Four questions you answer about yourself rather than about a scenario, and your own answers are the finding. Then the pattern that makes praise look useless and criticism look like it works, whatever you actually do.
Read