TST

Watch an AI agent work.

Pick a task and press play. You will see what an agent does at each step, the rules it runs under, and where a person stays in control. At the review step, that person is you.

Scripted walkthrough with sample data. No AI is running here and nothing is sent.

A customer email comes in. The agent sorts it and drafts the reply.

What came in

New email

Can you come out this week?

From
Alex R. <alex@example.com>
To
Shared inbox

Hi, our water heater started leaking last night. Do you have anything open Thursday or Friday? Also, do you charge a call-out fee?

Thanks, Alex

What the agent prepared

Nothing yet. The agent has not started drafting.

The agent's working note

New message in the shared inbox. Starting.

Step 1 of 6: Trigger

A new email lands in the shared inbox. That is the trigger.

Key idea

Trigger

The event that starts the agent: a new email, a form, a date, a schedule. Nothing happens until one fires.

Activity log

  1. Email received from alex@example.com

Short on time? The 38-second version

The inbox example from start to finish. There is no sound, every word is on screen, and the six stages below say the same thing in text.

The six stages, in plain words

Every task in the walkthrough follows the same six stages. Once you can see them, you can judge any agent someone offers to build you.

  1. Trigger

    Trigger. The event that starts the agent: a new email, a form, a date, a schedule. Nothing happens until one fires.

  2. Read

    Context. The agent looks things up in the tools you already use. It can only be as good as what it is allowed to read.

  3. Decide

    Guardrails. Rules you set in advance: what the agent may do, what it may never do, and when it has to stop and ask.

  4. Draft

    Draft, do not send. The agent prepares the work but does not act on it yet. A draft costs nothing to throw away.

  5. Review

    Human in the loop. A person approves, edits or rejects the draft. You can loosen this later, once the agent has earned it.

  6. Act and log

    Audit log. The agent does the task inside your tool and writes down exactly what it did, so anyone can check later.

What this page is, and is not

This is a scripted walkthrough with sample data. No AI model is running on this page and nothing you click is sent anywhere. It shows the shape of the work, not a result we are promising.

A real build connects to your own tools and runs under your own rules. The four tasks here are the examples from the AI Agent Starter: one workflow, fixed scope, 1 to 3 weeks. To see which of your own tasks is the best one to start with, try the automation finder.