Stop running payroll. Start supervising it.
The tasks your team redoes every month can become a script you write once and an agent that runs on schedule. One hour watching that happen, with the real product on screen.
From a team that has processed over R$2 billion in payroll with AI.
Two quick steps. The session link arrives in your work email.
Date
Tuesday, October 6, 2026, at 11am (Brasília time)
Format
Live webinar, about one hour, with questions at the end
Who presents
Tako's product and business teams
Who it's for
Payroll, People Ops and HR teams already using AI day to day
Pulling a simple report out of the system turns into an engineering project. The expiring vacation alert depends on someone remembering. Real payroll cost against budget lives somewhere else and arrives late. None of it is hard: it's repetitive, which is why nobody documents it. What we've seen in the months since the MCP launched is that these are exactly the tasks that leave the team's hands first. Not as an IT project, as a saved conversation.
No concept slides. Each one starts the way it lands on your desk: a messy file, a report request, a comp cycle spreadsheet.
Commissions, profit sharing, health plan co-payments. The file goes in as it arrived, the MCP matches it against Tako's records, shows what it's about to write, flags the off-pattern value and asks before moving on. Bulk entry stopped being the risky part of the month.
Overtime by bracket, cost per cost center, headcount over the last twelve months. You describe the cut and the layout once, including that specific format a client or an auditor demands from you, and get the file ready to forward.
Salary and role changed in bulk, from the spreadsheet the committee approved. Anyone above the salary policy ceiling doesn't slip into the batch: they come out of it and become an approval request, in the workflow the platform already has.
The question Finance asks every month. The MCP pulls actual cost, employees and contractors, and crosses it with the plan that lives in Drive, in Notion or in the FP&A spreadsheet. Out comes a chart by department, with the overrun pointed at.
Vacation about to expire, probation period ending, an onboarding stuck without a document. An agent wakes up on schedule, checks the records and posts to Slack. You only step in when there's something to decide.
The person who knows how writes the script once and gives it a name. After that, anyone in the team runs the same task the same way, without depending on whoever wrote it. The knowledge leaves one head and becomes something the company owns.
A skill is a saved script. You get the steps of a task right once, give it a name, and from then on, for anyone on the team, you just call it by name. An agent is the same thing without you asking: it wakes up on schedule, runs the check and only comes to you when there's a decision to make.
In the webinar both are built live, from scratch, in front of you. No library prepared the night before. The point is that you leave with a skill in mind to write that same week.
Payroll isn't left exposed. Every write follows the same six steps, and that holds for a skill you wrote yourself:
You access the AI with the same Tako login. No permission to edit means the AI can't edit either.
It reads Tako's records before proposing anything.
If something looks off-pattern, it stops and hands the decision back to you.
It runs one case before applying the whole batch.
Contract changes go through the same approval steps you already have.
Every edit is recorded under your name, with date and time.
Tako's structured records are what make this trustworthy. The spreadsheet that comes from the team always has an error in it. Tako is the source of truth that catches it.
Claude (our recommendation, faster and with more features), ChatGPT via Codex and Cursor. Use the desktop app, not the browser. Access uses your Tako login through Google SSO, and today it's open to admin profiles. The MCP is in beta and grows every week.
“It's a very good payroll system, different from the others I've worked with. I've saved a lot of time closing payroll.”
“We implemented Tako recently and it was well worth it.”
More than R$2 billion in payroll processed by Tako.
It is, and the overlap is small. June showed what the MCP does. This session shows what you build with it for the month-to-month: repeated tasks become a skill, and a skill becomes an agent that runs on its own. Bulk entry shows up again, but as the starting point, not the climax.
No. If you can explain the task to a new person on the team, you can write a skill. It's plain text, not code.
It is. The MCP works on top of Tako's records, so what you see on screen assumes payroll running here. Take it for what the session makes clear: the kind of task that stops existing once payroll data is reachable from inside the AI tool.
Claude, ChatGPT via Codex or Cursor, in the desktop app. Claude is the recommendation. Usage draws on the tokens of your own plan in that tool.
Under your name. You access the MCP with your Tako login, so every edit is named, with date and time.
Yes. Everyone who registers gets the recording link by email, even without attending live. But the Q&A at the end only happens live.
The session is live, with time for questions at the end. If you're already a customer, you leave with a skill in mind to write that same week. If you're not, you see payroll working a different way.
WhatsApp