DICA 1 · Diagnosis: the Client Map|INEMA.CLUB-PRO

Lesson 7 · 18 min

Risk, data and human dependency before you recommend

The candidate you picked is not ready to go into the report yet. First, four questions: how much can it do on its own, what personal data passes through it, who answers if it makes a mistake, and whether the channel is the official WhatsApp. In a small business, WhatsApp sits in the middle of almost everything, and its rules changed in 2025 and 2026.

Start the lesson
A clinic receptionist helps a patient at the front desk; beside them, a closed file cabinet.

In this lesson you will

  1. 1The 3 automation levels
  2. 2The minimum LGPD (Brazil's data protection law) for a small business
  3. 3The current WhatsApp Business rules
  4. 4The 4-question checklist before you recommend

Goal: check the chosen candidate for automation level, personal data, who answers if it makes a mistake, and WhatsApp rules before you recommend it.

01The 3 automation levels

The more risk and judgment involved, the lower the level. Click each one.

Click each block in the diagram to see what it does.

Showing where the person still approves answers the owner's fear before they ask.

02Data and channel: what holds today

Rules with dates and sources. This is not legal advice: when in doubt, consult a specialist.

1. Personal data

A customer's name, phone number, and conversation are personal data. Health data is sensitive data (LGPD — Brazil's data protection law, Law 13.709/2018, art. 5, II): at a clinic, almost every patient message is sensitive.

Sensitive data can only be processed under the cases in art. 11, such as specific and highlighted consent. And the person can ask for human review of a decision made only by a machine (art. 20).

03Template: checklist before recommending

One sheet for the candidate chosen in lesson 6. Replace what is between < > with your own.

checklist-risco.md
# Risk checklist — <candidate> — <business> — <date>

1. Automation level: <1 / 2 / 3>
   Why: <what happens if it gets it wrong; who approves>

2. Personal data involved:
   - Types: <name, phone, conversation, health data, financial>
   - Is any of it sensitive (health, religion, biometrics...)? <yes/no: which>
   - Basis for processing: <consent / contract / other; confirm with a specialist>
   - Vendor that receives the data: <which tool; where it stores it; does it use it for training?>

3. Who answers if it gets it wrong:
   - Person: <name/role>
   - How the customer asks for a person: <word, button, hours>
   - Human review of automated decision: <how>

4. Channel:
   - WhatsApp: <official API / official provider / unofficial / not used>
   - Messages outside the 24-hour window? <yes/no: which templates>
   - Did the customer agree to receive messages? <how it was recorded>
   - Does the assistant stay within the business scope (not free conversation)? <yes/no>

Conclusion: <recommend as is / recommend at level 1 / do not recommend now> — <reason>
Sources checked on <date>: <links to the rules>
Example (fictional)
# Risk checklist — after-hours questions on WhatsApp — Clínica Vila Serena (fictional)

1. Level: 1 in the first weeks (the front desk approves in the morning the replies prepared at night)
2. Data: name, phone number, conversation; health data shows up (pain, surgery) → sensitive
   Basis for processing: specific consent from the patient or another case under art. 11; confirm with a specialist
   High risk (ANPD Res. 2/2022, art. 4)? assess with a specialist: there is sensitive data and new technology
   Vendor: <to be defined>; check where it stores data and whether it uses it for training
3. Who replies: the receptionist; the owner in exceptions; R02 calls the physical therapist
4. Channel: today it's regular WhatsApp on the front desk's phone; to automate, only the official API
   Scope: hours, insurance plans, initial assessment; no free-form conversation

Conclusion: recommend at level 1, after writing the rules (lesson 5)
Simulation: illustrative example
  • The 4 questions have answers, not "to be seen."
  • The channel is marked as official or unofficial.
  • If there is sensitive data, it says which kind and who confirms the basis for processing.
  • The sources have the date you checked them.

04Practice: the candidate's checklist

Use the candidate recommended in your lesson 6 matrix.

E1 · 4-question checklist

20 min

Fill in the risk checklist for the recommended candidate at Vila Serena (fictional).

Done when: the 4 questions have answers, the channel is marked official or not, and the conclusion states the automation level with the reason.

E2 · Check the sources today

10 min

Open the WhatsApp Business pricing page and Resolution CD/ANPD No. 2/2022 (links at the end) and note in the checklist the date you checked and whether anything changed.

Done when: the checklist has both sources with today's date and one "changed / did not change" line.

05Test what you learned

Question 1 of 4

An assistant will prepare payment reminder messages. What level does it start at?skill

06Summary

Level 1 firstThe AI prepares, the person approves. Level 3 only for what carries no risk.
Health data is sensitiveLGPD arts. 5 and 11; small businesses have simplified rules, but AI with sensitive data calls for extra care.
WhatsApp, with datesPer message since July 1, 2025; general-purpose chatbots banned since January 15, 2026; 24-hour window and consent to receive messages.

Going further

Previous lessonNext lesson