Module · Orientation

Day-one automations

Lesson 1 of 9 · 3 min

Automation in sales and marketing is not one thing. At one end it is a rule so simple it feels like cheating: when this happens, do that. At the other it is an AI holding a phone conversation with a stranger and updating your CRM afterwards.

Both are built in the same place, out of the same pieces. Here are five, arranged from the simplest to the most ambitious, so you can see the range before you learn any of it.

The five

The problem	The automation
A lead arrives and sits there. Somebody has to notice, decide who owns it, and write the task. Usually the next morning.	A trigger on new contacts creates the task immediately, assigned by a rule. Two nodes.
A deal moves and nobody is told. The owner finds out at the weekly meeting, or when the customer asks.	A trigger on the deal record sends the email the moment the stage changes. Two nodes — and it is the next lesson.
The same six questions, every day, on WhatsApp. Returns window, warranty, where is my order. Somebody answers them all, badly, between other work.	An AI agent that searches your own policy documents and replies on the same thread. Four nodes.
Meeting notes that never become tasks. Everyone agrees to things and the notes go in a folder.	An AI reads the notes after the call and creates the tasks in bulk. Three nodes on the end of a branch.
A list of leads nobody has called. They age, they go cold, and calling them all is nobody's favourite afternoon.	A voice agent calls them, has a real conversation, qualifies them, and writes what it learned to the CRM.

A lead arrives and sits there. Somebody has to notice, decide who owns it, and write the task. Usually the next morning.

The automation: A trigger on new contacts creates the task immediately, assigned by a rule. Two nodes.

A deal moves and nobody is told. The owner finds out at the weekly meeting, or when the customer asks.

The automation: A trigger on the deal record sends the email the moment the stage changes. Two nodes — and it is the next lesson.

The same six questions, every day, on WhatsApp. Returns window, warranty, where is my order. Somebody answers them all, badly, between other work.

The automation: An AI agent that searches your own policy documents and replies on the same thread. Four nodes.

Meeting notes that never become tasks. Everyone agrees to things and the notes go in a folder.

The automation: An AI reads the notes after the call and creates the tasks in bulk. Three nodes on the end of a branch.

A list of leads nobody has called. They age, they go cold, and calling them all is nobody's favourite afternoon.

The automation: A voice agent calls them, has a real conversation, qualifies them, and writes what it learned to the CRM.

Two of those are rules. Three of them think. The first two you could build today with what is in the next module; the last three are the second half of this course.

What they have in common

Every automation in this course, including the ambitious ones, has the same three parts.

Something starts it. A record changed, a message arrived, a clock struck seven, somebody clicked a button. This is a trigger, and it is the first thing you choose.
Something happens. One step or twenty. A lookup, a decision, an email, a model, a phone call.
Something is different afterwards. A task exists, a message was received, a field was updated. If nothing is different, you have built a very reliable way of doing nothing.

The third part is the one to hold on to. It is the honest test of whether an automation is worth building, and it is also — as most of this course turns out to be about — the thing you should check after every run. A workflow that finished is not the same as a workflow that did something.

Choosing the first one to build

The instinct is to start with the impressive one. Do not. The first automation you build should be the one you would notice if it stopped — because that is the one you will maintain, and maintenance is what separates an automation from a demo.

Pick something that happens often. Once a day beats once a quarter. You will learn faster and you will notice problems sooner.
Pick something with an obvious wrong answer. An email that goes to the wrong person is visible. A subtly wrong summary is not, and you are not ready to debug that yet.
Pick something you currently do by hand. Then you already know the rules, and the workflow is a transcription rather than a design problem.

What to expect from this course

The next two lessons are watching rather than building — one rule-based automation and one AI agent, running. After that the Builder, and then every node in turn.

One thing worth saying at the start, because it shapes everything after it: most of what goes wrong in automation does not look like an error. Things complete. Runs are green. Nothing is red. A large part of this course is learning where to look instead — and it is why nearly every lesson ends with what breaks rather than with what works.

Next: Your first automation — two nodes, one deal, and an email that nobody sent.