Skip to content
beatriz.← Discernia
Beatriz AI Coach · by Discernia
beatriz.

Learning that stays.

The AI that teaches AI.

You don't need another AI course. You need a teacher of your own — by voice and text, with a whiteboard that comes to life on the spot for your case.

Start your pathAsk Beatriz
Beatriz introduces herself — Beatriz introduces herself in just over a minute: the board built on the spot, the plan, practice outside the platform and the autonomy test. Narrated and captioned in Portuguese; the full transcript is below.
Read the transcript

Hi. I'm Beatriz. An artificial intelligence — and your AI teacher. You already use AI every day. What no one taught you was how to use it with judgment — and how to stay good when it isn't around. A study with students measured this. With the AI on, performance went up forty-eight percent. With it off, it dropped seventeen — below those who never used it. The version with pedagogical safeguards had no such drop. That is the one I was built on. Here, every lesson is created on the spot, for your case. You speak or write; I build the whiteboard as I answer — and ask you back. Before teaching, I listen. An intake evaluation sets your path and your pace. The plan stays visible: lessons, tasks, deadlines. And I remember last week's mistake. Most of your learning happens outside of here, in your real work. That's why I send you off to practice — and wait for you to come back. Every three weeks, a checkpoint. At the end, the autonomy test: I step out of the picture, and we measure what stayed. Your history is yours alone. And your company's confidential data stays where it belongs: in its approved AI — I teach you to work that way. I don't do it for you. I enable you to do it. My success is you needing me less and less. Learning that stays. Start your path.

What stays when the AI leaves?
+48% / −17%
With the AI on, student performance rose 48%. With it off, it fell 17% — below those who never used it. The group with pedagogical safeguards had no drop.
ⓘ Source
Bastani et al., “Generative AI without guardrails can harm learning”, PNAS 122 (2025) — Wharton

Beatriz was built on the second condition: the one that teaches you to do, not the one that does it for you.

“I don't do it for you. I enable you to do it.”

The difference
01

Develops competence that stays

Every session ends with a task to apply in your real work. The next one starts from it.

02

No dependency effect

The weekly quota pushes you toward practice. And at the end, the autonomy test is taken with Beatriz out of the picture.

03

Proof, not perception

What you learned is measured with the AI switched off — by the LUMEN standard, published openly.

How it works

Voice and text in real time

You speak or write. Beatriz answers while building the whiteboard — and asks questions back.

Interactive whiteboard

Diagrams, steps and examples appear beside the chat, for your case.

Diagnostic intake evaluation

Before teaching, she listens. Your path and pace come from there.

Visible plan on a timeline

Lessons, tasks and deadlines in a development plan you can follow.

Continuous memory

She remembers last week's mistake — and starts from it.

Tailor-made lessons

Content is created on the spot, including about the prompt or the agent you are creating.

The profiles

Two profiles, with twelve competencies each. The profile is set at enrollment; the emphasis of the lessons adjusts to your context — whether or not you lead people, company size, maturity of the tools; or whether you are about to start a business, already run one, or sell your hours or content.

Profile 1 · AI at corporate work — professionals and team managers (12 competencies)

01How AI works, gets it right and gets it wrong in everyday work
Anticipate where AI fails predictably — plausible text with no grounding, information that gets lost in the middle of a long document, uneven quality — and choose the instrument each task calls for.
02Asking well: instructions, company context and material you reuse
Instruct the AI with the context that exists only inside the company — process, policy, approved format, the department's voice —, iterate until it works and turn what worked into a reusable instruction.
03Check before you deliver: the deliverable carries your name
Verify fact, number, quote, calculation, source and deadline before the output becomes a deliverable, with rigor proportional to the cost of the error.
04Company data in the right tool: confidentiality, corporate account and shadow use
Take care of what goes into the prompt and where it is typed: corporate account, what can and cannot be pasted, unapproved tools, retention and training on what you send.
05Personal data, LGPD and the internal rule governing AI use
Recognize personal and sensitive data in the material, apply purpose limitation and minimization and operate within the internal policy — what requires approval, what gets recorded, whom to escalate to.
06Documents, spreadsheets and analysis: office work with AI
Summarize, compare, extract, structure and calculate on real work material — and know when the result needs to be validated against the original source.
07Work communication with AI: announcements, minutes, dashboards and a Q&A assistant
Use AI in work channels with notice and consent settled before recording — and with a clear separation between what was said and what the tool inferred.
08Automate your own routine without coding
Build, with approved tools and no code, assistants and workflows for the recurring part of the job — defining the data source, minimum permissions, a review step and what never runs without supervision.
09Your own judgment: using AI without outsourcing the competence
Recognize when a well-written answer is wrong, disagree with arguments and decide when not to use it — preserving the skill that sustains the role.
10Bringing AI into the team's work: agree on the rule, redesign the step and show the gain
Find the step in the team's workflow where AI actually solves something, agree in writing on what is allowed and who reviews, record the baseline and demonstrate the gain — or end the use that did not hold up.
11Responsible use: bias, transparency and decisions that affect people
Recognize when using AI stops being neutral — candidate screening, customer service, published content, other people's image and voice — and act: disclose, require a real human decision or do not use it.
12Your work from here on: what AI changes in the role and the career
Read the effect of AI on your own occupation — what gets absorbed, what gains value, which competence to sustain — and turn it into a concrete plan, without panic or denial.

Profile 2 · Starting a business and working for yourself (12 competencies)

01AI without the traps: where it delivers, where it fails and which tool each problem calls for
Anticipate, before using it, whether AI will help or make the task worse — and choose the instrument: a one-off question, a reusable instruction, an agent, an automation, a simple system or an off-the-shelf solution.
02Prompts that work: instructions, your business context and materials you reuse
Instruct the AI with your business context — offer, price, audience, brand voice, recurring objections —, iterate until it works and keep a library of reusable instructions.
03Your name on the deliverable: check what the AI produced before the client sees it
Check fact, number, price, deadline, promise and image before they reach the client: with no reviewer and no legal department, the one who checks is the one who signs.
04Building without coding: skills, agents, a web page and simple systems for your business
Build structured instructions, agents, a web page and simple systems for your own business — and be able to describe what each one does, which exceptions it handles and where it breaks.
05Autopilot: automating the administrative and financial routine
Build workflows that run without you present — billing, reconciliation, reminders, follow-up, order triage — with error handling and a defined supervision point.
06Plugging AI into your day: calendar, messages, spreadsheets and connectors
Connect the AI to your calendar, messages, spreadsheets and business systems with the smallest permission scope that gets the job done and sensitive actions under confirmation.
07Customer service and scheduling with AI without losing the customer
Run AI-assisted customer service with a source of truth, limits on what can be promised, a handoff rule to a person and sample review — what the assistant states counts as your word.
08Marketing with AI: content, social media and customers coming in
Produce and maintain a presence that generates demand — topics, copy, images, calendar, publishing — with brand voice, review before publishing and an honest reading of what brought a contact, a quote request or a sale.
09Your face and your voice with AI: avatar, video and cloning with consent
Produce video and audio with a cloned avatar, image and voice while settling consent, ownership and labeling of the material as synthetic — and know when the replica strengthens the brand and when it erodes it.
10LGPD in practice: customer, lead and payment data when you are the one responsible
Handle customer, lead and payment data correctly: recognize personal data, legal basis and consent, minimize, choose the tool by its data-processing regime and respond to data-subject requests.
11Rules of the game: advertising, professional bodies and what you can promise
Recognize what governs AI use in your line of work — professional body, advertising, consumer law, image rights and authorship — and apply it: what requires labeling and human review, what cannot be promised, when to consult a lawyer.
12Does the math work? AI cost, measured gain and what is worth keeping
Estimate what a solution will consume before building it, record the baseline, measure the result in time, cost or revenue and cut what does not pay for itself by a criterion set in advance.

Rules that apply to both profiles: a gain obtained by skipping the data or confidentiality rule is not performance; “the AI did it” does not transfer the responsibility of the one who signs; content about laws and regulations is educational, never legal advice.

Governs and proves
01

Guardians at every turn

Four layers of guardians and a judge at every turn keep Beatriz in her role: to teach, not to do it for you.

02

Checkpoints every 3 weeks

Small tests of real practice, in your work, with feedback.

03

Autonomy test

At the end, with Beatriz out of the picture and a 30-to-50-day interval: we measure what stayed.

◆ Measured by the LUMEN standard
Ask Beatriz

Three questions everyone asks — answered by her. The video versions are in production.

?What is a lesson with you like?
You choose where to start — or tell me the real problem on your desk. We talk by voice or by text; I build the whiteboard as I answer and ask questions back until you can do it on your own. At the end, a task to apply outside of here. In the next session, I start from it.
?How much does it cost — and what if I use more?
You buy a package with a clear usage ceiling and an estimate of hours per week: two on Autonomia, four on Domínio. The week's quota does not roll over — when it's used up, you practice and come back. If you need more, there are extra blocks at a fixed price, no surprises. And if you complete the weaning evaluation on time, the last two monthly installments are canceled. Prices are on the enrollment page.
?Is my data safe?
Your history is isolated: no student sees another's. Your voice becomes text in your own browser — I don't record audio. And your company's confidential data does not come in here: I teach you to work in its approved AI and to bring an anonymized version to practice with me.

In the next phase, this button opens Beatriz live, scoped strictly to these three questions.

Plans

You buy a usage package with a clear ceiling. Hours are a consumption estimate — not the billing unit. Individual enrollment online, with prices on the enrollment page; companies and institutions by proposal.

Autonomia

About 2 hours a week · estimated 48 hours in the package

  • 8 monthly installments
  • Path, whiteboard, plan, checkpoints and autonomy test
  • Completion discount: complete the weaning evaluation on time and the last two installments are canceled
  • Extra blocks at a fixed price, if you need them
See prices and enroll

Domínio

About 4 hours a week · estimated 96 hours in the package

  • 8 monthly installments
  • Everything in Autonomia, with twice the guided practice
  • Completion discount: complete the weaning evaluation on time and the last two installments are canceled
  • Extra blocks at a fixed price, if you need them
See prices and enroll
  • The weekly quota does not roll over: when it's used up, you practice in your work and come back the following week.
  • The package is measured in consumption (tokens). The hour estimate is for your planning — actual consumption varies with the use of voice, whiteboard and lesson generation.
  • Payment by credit card. 7-day right of withdrawal, under the Brazilian Consumer Defense Code.
Security by design

Isolation per student

No student sees another's history. For companies, the client dashboard shows the cohort's progress, never the conversation.

Your voice is not recorded

Speech becomes text in your own browser; the server receives only text. Caveat: the browser's speech recognition may send audio to the browser's maker, under its own policy.

Approved AI for real data

Your company's confidential data does not come in here. Beatriz teaches you to work in your company's approved AI and to bring an anonymized version to practice.

LGPD and auditing

Declared legal basis, minimal data, guardrails and sample auditing of sessions.

For companies

Cohorts by area, with proof of value in the weaning evaluation. Expansion by area and renewal based on what stayed — not on attendance.

  • Scope → diagnostic → paths under way → dashboard tracking → weaning evaluation
  • Direct contract and invoicing; cohort terms by proposal
  • Client dashboard at cohort level
Request a company proposal

For educational institutions

Two doors — students and faculty. Non-exclusive partnership, in five formats: pilot, co-branded module, continuing education, faculty training and research stream.

Talk about a partnership

What Beatriz doesn't do

  • She is not a K-12 tutor.
  • She does not give legal, medical or financial advice.
  • She is not a perimeter for real corporate data.
  • She does not sell unlimited hours — and does not promise to “master X in 7 days”.

Start your path.

A teacher of your own. A whiteboard that comes to life on the spot. Proof that it stayed.

Start your path