Decisions with evidence. Agents with numbers.
GenAI BR's consulting takes Cinco's method inside your company: primary sources, measurement before opinion, risks in writing. Small steps, each with a fixed price, and each one leaves you with something useful — even if you stop there.
It is not transformation consulting, a two-year roadmap or developers by the hour. How to start↓
The steps
Each step is paid, has a fixed price in the proposal and ends with something in your hands. You can come in through any of them; the first two are the doors: the Briefing answers the decision-maker's question, the Workshop the team's.
- 1
Briefing
One question your company has about AI.
- Ten pages at most, with a source for every claim
- One hour of conversation with the person who read
- The appendix with the sources, so the team can carry on alone
- A one-page update, thirty days later
- 10 working days
- 2
Workshop
Half a day or a day with your team, up to 20 people, working on your own tasks.
- Before: 15 to 20 real tasks from the area, anonymized — without them, there is no workshop
- On the day: one third evidence, two thirds hands-on, with the tools the company has already approved
- After: a document with what AI solves today, what it does not, the risks and a 30-day plan
- Thirty days later: one hour to see what the team actually adopted
- half a day or a day
- 3
Diagnostic
One process or one team, seen from the inside.
It is the proof of concept — paid, short, and yours.
- The set of real cases for evaluation — it is yours, and works with any vendor
- Where AI fits and where it does not, with the why
- The estimated cost per case
- A yes or a no, in writing
- 2 to 3 weeks
- 4
Pilot
Building with brakes, in production.
- The success criterion, written before starting
- In production (the agent in the process, or the team working the new way), with human approval on what is sensitive and a record of every action
- A weekly dashboard with the agreed numbers: accuracy, interventions, cost, lead time
- If the number does not come, the pilot stops — and that is a result too
- 12 weeks
- 5
Follow-up
What comes after.
- A monthly dashboard
- Re-evaluation against your set of cases when the model, the price or the vendor changes
- One hour a month with the person who read
- monthly
On what
Two subjects, with the same steps. The house starts with engineering, where Cinco's reader already is and where the proof is already public.
- Open · every step
Software engineering with AI
For development teams that already use coding assistants and want measured results, not impressions. The core is specification-driven development: the specification as the source of truth, code as a derivative, human gates and traceability — and the metrics of tool use, before and after.
- Our team writes code with AI. Is it faster or just busier?
- How do we put specifications into the flow without turning it into bureaucracy?
- What do we measure when half the code is generated?
- Briefing and Workshop now · Diagnostic and Pilot on the Roadmap
Agents in processes
For operations with volume and rules: customer service, document triage, reconciliation, onboarding. It is Ofício's legacy: the agent with a record of every action, human approval on what is sensitive and continuous evaluation against real cases. The Workshop for business areas — legal, HR, customer service, finance — triages the area's tasks.
- Is our customer service ready for an agent that acts — or only for one that reads?
- Which legal tasks does AI solve today, and which not?
- What would it cost, per case, to automate document triage?
- Oct 2026Consulting open, starting with AI-assisted software engineering: workshop, diagnostic, pilot and follow-up
- Oct 2026Workshops for business areas: triaging the area's tasks
- undatedDiagnostic and Pilot for agents in processes: customer service, document triage, reconciliationBecomes planned when: the first engineering pilot ends, with a case the client allows us to show
How we work
- 1Map
Inside the process or the team: where the rules are clear, where they are not, where a mistake costs.
- 2Evaluate before building
The set of real cases comes first. Without it, nothing gets built — and with it you can compare any vendor, not only the one we recommend.
- 3Build with brakes
Human approval on everything sensitive. The brakes loosen only with numbers.
- 4Operate and measure
Weekly dashboard: accuracy, human interventions, cost per case. Open to the client, always.
House rules
A house that publishes about AI and sells AI consulting has to say where it makes money. These rules are the public side of the Method's register of interests. The register is on the Method page.
- No vendor commissions
- No AI vendor pays the house for referrals, resale or licenses. When we recommend a tool, we gain nothing from the choice.
- Spectrolite is ours, and that is written down
- If your problem has to do with Spectrolite, the relationship is on the first page of the proposal and of the document, with at least one alternative that is not the house's.
- What is yours stays yours
- Data, code and the set of evaluation cases belong to the client. The house keeps the method.
- Clients do not become stories
- Nothing the consulting sees becomes Cinco. A case is only published with written permission, and with the relationship declared.
- Confidential from the first call
- A non-disclosure agreement before the first call; a data processing agreement (Brazil's LGPD) when there is personal data; confidential data only in tools the company has approved.
- Fixed price, never by the hour
- Each step has a fixed price in the proposal. Working with AI and charging by the hour rewards slowness.
Who does it
GenAI BR's founder. Works with every client personally.
The person who reads for Cinco every week is the one who does the consulting. There is no junior team behind a partner who shows up at the meeting.
Author of the Spec-Driven AI framework — the specification as the source of truth, code as a derivative, human gates and traceability — and first author of papers accepted at ICSME 2026 and ICTAI 2026. Gives talks and classes on AI-assisted software engineering, agents and specification-driven development, after some seven years in data science and AI.
The best sample of how the house works is free: Cinco, the Archive and the Method. Spectrolite is built the same way the consulting teaches. Client cases will appear here with permission.
Slots and price
- 2
- 2
- 1
Slots are few on purpose: the consulting cannot delay Tuesday's edition. When they are taken, the reply brings the next date.
A fixed price per step, in the proposal, 48 hours after the first call.
The Workshop is priced per workshop, not per person, up to 20 people. In person, travel is extra.
Questions
- Do you implement?
- Yes, in the Pilot. The Briefing, the Workshop and the Diagnostic end at the recommendation; the Pilot builds, with brakes, and measures.
- Do you do free proofs of concept?
- No. The Diagnostic is the proof of concept: short, paid, and it leaves you the set of cases to evaluate any vendor — including one other than the recommended.
- Do you run open courses or talks?
- Open courses, no. A closed workshop for your team, yes. João gives talks on his own; they are not a product of the house.
- Remote or in person?
- Both. In person, travel is extra.
- Can part of the payment depend on the result?
- In the Pilot, it can be agreed: the last instalment depends on the success criterion written at the start.
- What if the best answer is Spectrolite?
- It is written as such, on the first page, with at least one alternative that is not the house's. You decide.
How to start
Your question in one sentence.
Write to the address below. I reply in about two working days. The first call is thirty minutes and costs nothing; the proposal, with fixed scope and price, comes in 48 hours. No form: an e-mail is enough.
Write to ola@genaibr.ai→Get the next Cinco
Tuesday, 7am BRT, in your inbox. Five items, with sources. No daily newsletter, no promotions, and unsubscribing is one click.