Agents in production
Task-specific agents wired into the tools you already run, with access control, logging and a defined path for the cases they should not handle alone.
Applied artificial intelligence, São Paulo
NASCIA AI designs, builds and runs agents, evaluation harnesses and integrations inside the processes your organisation already uses every day. Not a demonstration: a system your own team can operate, audit and extend.
The same four steps, in the same order, on every engagement.
What we build
An agent is easy to demonstrate and hard to keep running. The difference is not the model. It is everything around it: the data it is allowed to see, the evaluation that tells you when it degrades, the fallback when it fails and the person who owns the outcome.
Task-specific agents wired into the tools you already run, with access control, logging and a defined path for the cases they should not handle alone.
A test bed that tells a real improvement from a lucky run, so a model or prompt change never ships on a hunch.
Illustration with simulated data
The plumbing between the model and your systems of record: ERP, CRM, ticketing, finance and documents, built to be audited.
Practical programmes for companies, schools and foundations, taught by someone who builds these systems as well as writes about them.
In production
Four of our live builds, each with one figure, its source and its date, and one confidential enterprise system.
Education
More than 15,000 educators a month.
Source: ai-teachers.pro and mrnascimento.com, read on 24 September 2026.
Read the case study: AI Teachers
Finance
415 tests passed in one adversarial run against the output boundary, 11 files.
Source: Portfolio Horizon claims register, run of 31 July 2026.
Read the case study: Portfolio Horizon
Education
12 stages in the AI orchestration that maps each topic to UK careers.
Source: Careers Studio landing page, read on 24 September 2026.
Read the case study: Careers Studio
Education
480 IGCSE past-paper PDFs in a protected archive.
Source: mrnascimento.academy home page, read on 24 September 2026.
Read the case study: Mr Nascimento Academy
Finance
For a privately held group: the first line of the monthly income statement (DRE) produced automatically from raw ERP data, validated against six months of historical data and accepted in writing by the executive sponsor.
Read the case study: Enterprise finance agentsConfidential client
Delivered through more than ten versioned releases, with more than 60 pages of documentation.
Evaluation
Two versions of a model are scored on the same evaluation set. Move the slider and watch what the sample size does to the answer.
Illustration with simulated data
Accuracy, on an axis from 50% to 100%
At 100 examples the two intervals overlap, so this picture alone cannot tell an improvement from luck. Overlap does not show that the versions perform the same.
Difference: 4.0 percentage points
| Version A | Version B | |
|---|---|---|
| Accuracy | 82.0% (82/ | 86.0% (86/ |
| 95 per cent confidence interval (Wilson) | 73.3% to 88.3% | 77.9% to 91.5% |
| Interval width (percentage points) | 15.0 | 13.6 |
Each score carries a 95 per cent Wilson confidence interval. The chart shows uncertainty; it is not a significance test. A real comparison of two models on the same examples needs a paired analysis of their outcomes, which these simulated totals do not include.
The scores are generated for this demonstration and do not come from a client system.
How we work
The order is the method. Nothing is built before it is defined, nothing ships before it is measured, and nothing is handed over without the people who can keep it running.
Name the task, the owner and the measure of success before any model is chosen.
You receive
The written definition, with the data access scope and a go or no-go decision on building.
Build the evaluation harness first, so improvement can be told from luck.
You receive
The evaluation harness: a test set from real cases, metrics, baselines and a report that runs on every change.
Ship into the existing workflow, with fallbacks, logging and access control.
You receive
The integrated system in production, with its escalation paths, monitoring and audit trail.
Leave the runbook, the tests and the team that can operate it without us.
You receive
The runbook, the test suite with its evaluation history and the documentation in your repository, and a trained operating team.
The measure of the work is not the demonstration. It is whether the system is still running, still measured and still owned by your team after we have left.
Training
Programmes for companies, schools and foundations, in Portuguese and English, from executive briefings to hands-on workshops for engineering teams and training for teachers.
Leadership and engineering teams that have to decide what to build, what to buy and how to measure it.
Teachers and school leaders who need a working AI policy and classroom practice that holds up.
Teams that fund or run education and public services and need to judge AI proposals with evidence.
Founder

Founder, NASCIA AI
Author, mathematician, AI solutions developer, applied AI researcher, speaker and teacher.
Author of Generative AI for Teachers: A Practical Guide to Educational Technology (2025) and of Inteligência Artificial Generativa para Professores (Editora Dialética, 2026).
Mathematician. PhD candidate in Pure Mathematics at IME-USP, University of São Paulo, with the qualifying examination passed in 2026.
AI solutions developer who builds intelligent systems for large organisations and leads their AI implementation from first definition to handover.
Research and writing
A forensic study of answer leakage in agentic code repair, and opinion pieces on AI evaluation, education and public policy in the Brazilian and international press.
Study
Zenodo, author preprint, not peer reviewed. DOI 10.5281/zenodo.21286873 ·
A preprint by NASCIA's founder audits three of his own experiments, finds the expected output leaking through execution feedback, and sets out the controls a credible result needs.
Read the case studyThe AI Journal · · English
GovInsider · · English
EUobserver · · English
Questions
Plain answers on how an engagement runs, how we handle data and confidentiality, and what is out of scope.
One conversation is usually enough to tell whether an agent belongs there, what it should be measured against, and what it would take to keep it running.