1 · New model test
15 minFill in the test with three of your own deliverables.
Done when: each deliverable is marked "disappears" or "I get faster" and you wrote what stays with you.
Lesson 7 · 15 min
Human judgment: the new-model test and your customer's sensitive data
If a new model comes out and your work disappears, you were working the tool. If a new model comes out and you just get faster, you were working your judgment. In this lesson you run that test, learn to separate sensitive data before plugging AI into your customer's information, and learn to prospect before you have social proof.
Start the lesson
In this lesson you will
- 1The new-model test
- 2Draft the plan before running anything delicate
- 3Three data groups: open, restricted, local
- 4Prospect with a case video and a landing page
Goal: apply the new-model test to your work and sort a customer's data into three groups before plugging in the AI.
01Your customer's data in three groups
Case: an architecture firm with all its files on a local server (NAS). Click each group.
Click each block in the diagram to see what it does.
Give the client two options in writing: they own the data and are responsible for it, or they build it on your setup with clear rules.
02Judgment before tools
Four habits.
1. The test
Ask yourself: if a model comes out tomorrow that does what I do, does my work disappear or get faster?
If it disappears, you sell a tool. If it speeds up, you sell judgment and method.
2. Cooking
On a delicate topic (finance, sensitive data), spend hours on the plan before you execute.
Talk it through with an AI that has no context from the project, so you think from the outside, and only then build the plan with the agent that knows everything.
3. Separate
Before you connect the AI to the client's server, list the folders and mark each one: open, restricted or local.
Move the restricted data out of the folder the agent can see. Connecting first and separating later is too late.
4. Prospect
No testimonial yet? Record a short video explaining the case you handled and what changed for the client, without names.
Build a landing page with the video and a button for a 15-minute conversation. When you finish a project, ask for the testimonial.
03Templates to copy
The first one you answer yourself. The second goes to the client.
New model test
WHAT I DELIVER TODAY (3 items): 1. <deliverable> 2. <deliverable> 3. <deliverable> FOR EACH ONE: if tomorrow a new model does this on its own, [ ] it disappears [ ] I get faster WHAT STAYS WITH ME ON ANY MODEL: - <judgment I apply: what to review, what to refuse, the order of the pieces> - <documented method> NEXT STEP: turn 1 "disappears" item into a method (lesson 5).
Sorting the client's folders
CLIENT: <name> SERVER: <NAS / cloud / mixed> DATE: <dd/mm/yyyy> FOLDER | GROUP | REASON /propostas | open | commercial templates and proposals /projetos/<cliente-sigilo> | restricted | confidentiality agreement /financeiro/socios | restricted | partners' personal data /pesquisa | local | core of the business, needs AI CLIENT'S DECISION (signed): [ ] the client owns the data and is responsible for it; I apply the security [ ] built on my setup, with the rules above
04Practice
Close out the course with these three.
2 · A client's folders
30 minPick a client (or your own company) and sort the main folders into the three groups.
Done when: the table has at least 6 folders, each with a group and a reason.
3 · Prospecting with the case
40 minWrite the script for a video of up to 90 seconds about a case of yours: the problem, what you built, what changed.
Done when: the script has all three parts, no client names, and ends with the invitation to a 15-minute call.
05Lesson 7 on video
The lesson's mechanism explained in about two and a half minutes.
Lesson 7 on video · 2 min 24 sVideo page with script →
06Quick quiz
4 questions.
Question 1 of 4
You sell automations on a workflow platform. An agent comes out that does the same thing on its own, and your orders drop. What does this show?
Correct
If your work disappears with the new model, the value was in the tool. The way out is to sell the judgment and the method that still hold up on any model.Not that one
If your work disappears with the new model, the value was in the tool. The way out is to sell the judgment and the method that still hold up on any model.Review: Judgment before toolsQuestion 2 of 4
The office wants the agent working on the NAS files, where confidential projects are also stored. What is the first step?
Correct
Once connected, the agent already sees everything. Separating first, with the client's decision in writing, is what protects the business's goodwill.Not that one
Once connected, the agent already sees everything. Separating first, with the client's decision in writing, is what protects the business's goodwill.Review: Your customer's data in three groupsQuestion 3 of 4
In which case is it worth spending hours on the plan before executing?
Correct
A sensitive topic, with money and many sources, calls for judgment before execution. A simple, reversible task doesn't need hours of planning.Not that one
A sensitive topic, with money and many sources, calls for judgment before execution. A simple, reversible task doesn't need hours of planning.Review: Judgment before toolsQuestion 4 of 4
You don't have any testimonials yet. How do you prospect?
Correct
The case told as problem, solution and result is already convincing, without exposing the client. Making up a testimonial is lying; waiting costs you months.Not that one
The case told as problem, solution and result is already convincing, without exposing the client. Making up a testimonial is lying; waiting costs you months.Review: Judgment before toolsQuiz complete
You got {a} of {n} right. Get them all right and the lesson is marked as complete.Pick an option. The explanation shows up after you try.
07Summary
The testDoes a new model wipe out your work or speed it up?
Cook the planOn sensitive work, hours of planning before execution.
Three groupsOpen, restricted, local. Separate before you connect.
Order firstThen context and rules. That's when the AI does what it promises.