Today
Every area an island.
Data moves hand to hand, sheet to sheet, and the report arrives when it arrives.
For CEOs and directors of mid-sized companies: AI in production, with the accuracy target, price and deadline fixed before work starts.
On WhatsApp. I reply as soon as I can.
01The cost off the balance sheet
18hours per week on repeated tasksillustrative example of one week
Monday to Friday, the same tasks. Data copied from documents into spreadsheets, contracts checked by hand, the report every Friday.
Almost all of it follows the same rules. An invoice becomes a spreadsheet row. A common question gets a standard answer. An off-template clause becomes an alert.
Anything that follows rules becomes a process in the system. The routines are executed by the system, and the team checks what it sets aside for review.
If two or more of these happen at your company, it is worth a conversation. Whatever you mark goes straight into the WhatsApp message.
Mark what happens at your company.
What does this cost per year?
Illustrative example, 48 working weeks a year. Adjust to your team.
With 30% to 80% less time spent on these tasks, that would be 259 to 691 hours back per year, worth €5,180 to €13,820.
The 30% to 80% range comes from automations at KPMG (tax practice), in high-volume document workflows. The return for your case is estimated in the proposal, after the 30-minute call.
02Impasse
nearly 9 in 10
respondents say their company already uses AI regularly in at least one area.
McKinsey, The state of AI in 2026
37%
report that AI has had some effect on the company’s operating profit (EBIT).
McKinsey, The state of AI in 2026
about 6%
attribute 5% or more of EBIT to AI and say the value created was significant.
McKinsey, The state of AI in 2026
According to MIT NANDA, most AI systems in companies fail because of brittle workflows, because they do not learn from context and because they do not fit into day-to-day operations. What separates those who see a return, the study says, does not seem to be the quality of the model but how it is deployed. That is where engineering comes in.
In the same study, projects built with specialized partners were twice as likely to reach operation as those built only in-house.MIT NANDA, The GenAI Divide, Jul. 2025. Base: interviews with 52 organizations; results as reported by them.
In your company, the symptom usually goes by another name: the monthly report.
03Live department
Today
Data moves hand to hand, sheet to sheet, and the report arrives when it arrives.
1 of 4 · Support
Replies instantly, asks the first questions and books appointments, at any hour, following your business rules. Hands off to a person when needed and logs every contact in the CRM.
What changes: first-response time and logged contacts.
2 of 4 · Documents
Contracts, invoices and reports read and checked in minutes, every data point traced to its source page. The team reviews only exceptions.
What changes: checking hours and data-entry errors.
3 of 4 · Processes
Spreadsheets, emails and systems integrated and exchanging information, and the business numbers in an always up-to-date dashboard, with nobody building a spreadsheet.
What changes: close time and decisions with today’s number.
4 of 4 · Knowledge
Your team asks the way they would ask a colleague and gets the answer along with the internal document it came from.
What changes: time for new hires to learn the process, and repeated questions.
Whole department
Together, the four take a department from first contact to the report. CRM setup and team training are part of the project when the process calls for it.
↓Response time↑Team hours given back↓Data-entry errors↓Month-end close
Each system is built with your company’s real cases and has an accuracy target tested before delivery.
If this is for you, the 30-minute call is free and carries no obligation.
04The engineer

“At KPMG, I built AI systems for tax, legal and compliance, where a mistake is expensive.”
Leonardo Sá, AI engineer, in charge of MasterTech
I am a co-founder of Revisa Master, where I run several AI agents in production, used by paying clients every day.
I trained more than 1,600 professionals at KPMG. I studied at USP (University of São Paulo), Fatec and Université Lumière Lyon 2.
The same engineering, on a problem the size of your company.
05Method
From first contact to a working system. Scope, timeline, price and target written down before work starts.
Reply as soon as possible
On WhatsApp, you describe the department or task that weighs the most and agree on a free 30-minute video call.
30 minutes, by video, free
The engineer learns the process, where the cost sits and whether AI can solve it. If it can’t, you find out right there.
Price and target fixed before work starts
Scope, timeline, price and accuracy target defined before work starts. When the path is not yet clear, a paid diagnostic, credited toward the project.
2 to 6 weeks per project
Weekly deliveries you see working, built with your company’s real cases. From your team, a few hours a week to show the process and validate the deliveries.
In your accounts, documented
The system runs in your company’s accounts, with the team trained. Monthly maintenance is optional.
A 30-minute video call, free and with no obligation.
06Risk under control
The accuracy target is written into the proposal, next to the price and the timeline. The system is only considered delivered once it hits the target in tests with your company’s real cases.
The line is the accuracy target agreed in the proposal.
Whenever possible, the system shows where each answer came from, so your team can check it.
You pay for the project, not for a salary or by the hour. For companies with several requests during the year, there is a monthly package, with scope and target for each request.
07Return
The system and the code belong to your company, delivered documented and running in your accounts, so that any engineer can continue the work.
Delivered with documentation, so any engineer can continue the work.
Runs in your company’s accounts and, when needed, on your own infrastructure.
Your team comes out trained to take over the system day to day.
After delivery, you choose: your team takes over, with training and documentation, or MasterTech keeps maintaining and improving the system.
To take to the decision-makerScope, timeline, fixed price and estimated return in hours or cost, in a single proposal. If useful, the engineer joins a conversation with the decision-maker.
And your team goes back to the work it was hired to do.
Tell me on WhatsApp. I will reply and tell you honestly whether it can be automated and what the path would look like.
Leonardo Sá, engineer in charge
No obligation: if AI does not solve your case, you will know during the call.
It depends on what needs to be done. After the call you receive a proposal with a fixed price and timeline, before any commitment. Monthly costs of AI services, when there are any, are estimated in the proposal.
Yes. Companies without enough demand for a full-time AI team hire only the project they need, with scope and accuracy target agreed in the proposal. The proposal shows the estimated return in hours or cost, next to the price.
Every system has an agreed accuracy target and is tested with your company’s real cases. If the target is not hit in testing, it is only considered delivered once adjusted. Whenever possible, it shows where each answer came from, so your team can check it.
The agent introduces itself as an assistant, answers what it knows using your business rules and hands off to a person as soon as the conversation calls for it. You approve the answers before they go live and receive a summary of every interaction. The customer is served instantly instead of waiting until the next day.
A fixed-scope project takes 2 to 6 weeks, with weekly deliveries. From your team, a few hours a week to show the process and validate the deliveries with real cases.
In most cases, yes: WhatsApp, spreadsheets, email, and management, accounting and customer systems. If your system is already part of the task, the automation picks up where it stops: at whatever someone types, checks or copies by hand today. This is assessed on the first call.
Yes. A confidentiality agreement is signed before any access, with minimum access and use only for the agreed purpose, in line with the GDPR and Spain’s data-protection law (LOPDGDD). You know which AI services process the data, configured not to use it to train models. When needed, the system runs on your company’s infrastructure.
Yes. The system and the code belong to your company, delivered documented and running in your accounts, so that any engineer can continue the work.
The proposal comes with scope, timeline, fixed price and estimated return in hours or cost, ready to take to the decision-maker. If useful, the engineer joins a conversation with the decision-maker.
100% remote, with video meetings and a direct channel throughout the project. After delivery, you choose: your team takes over, with training and documentation, or MasterTech keeps maintaining and improving the system, with scope and target for each improvement.
Missing a question? Ask directly on WhatsApp (opens in a new tab)