About Pacific Machines
AI should start with how work really happens.
We are a small team in San Francisco building private work intelligence: software that learns from approved real work and shows where AI will pay off first.
- Based in
- San Francisco, California
- Company
- Reefo, Inc.
- Product
- Desktop app for macOS and Windows
- Contact
- iago@pacificmachines.ai
Why we exist
Most AI plans start from a guess.
The real workflow lives in the handoffs, exceptions, and copy-paste between tools. That is where we look.
Teams are asked where AI should go before anyone can see how the work is actually done. Process maps describe the ideal. Interviews describe what people remember.
We started Pacific Machines to make the work itself visible. It learns from normal days at the computer, finds the patterns and exceptions that repeat, and turns them into opportunities a team can inspect.
The person who does the work stays in charge of the record. The raw capture stays on their computer, and a person decides what moves forward.
What we believe
Four commitments behind the product.
They shape what we build, and what we leave out.
Quick to a first answer
Bring one workflow. In as little as a week, you see what to automate first and why.
Evidence before automation
Every opportunity shows the work behind it, so a team can inspect it before funding it.
Private by design
No screenshots, video, or audio. Sensitive text is redacted before it is stored, and the raw record stays on the computer.
People decide
Pacific Machines proposes. A person reviews the evidence and chooses what moves forward.
Team
Built by operators and engineers.
Three co-founders who have worked inside the operations AI is meant to improve.
Iago Maciel
Co-founder & CEO
Product and fintech operator. Built banking and payments products at BEES, AB InBev's B2B platform, and grew the cyber and insurtech business at EXA. Leads company direction and customer work.
Hans Preinfalk
Co-founder & CTO
Product engineer. Previously co-founded Nerdapp.ai, a vibe-coding platform used by more than 300,000 people across the Americas and Europe. Leads product and engineering.
Eduardo Schuch
Co-founder & COO
Early member of IBM Watson in Brazil, where he brought Watson to customers across Latin America. MIT MBA. Leads delivery, customer success, and operations.
FAQ
Questions, answered.
The short version. The product documentation has the rest.
What does Pacific Machines do?
It learns how work actually happens from approved, everyday activity on the computer. It finds the patterns and exceptions that repeat and turns that evidence into insights, opportunities, and reports your team can inspect.
Who is it for?
Operators who know where the work is heavy, and operational leaders who want evidence before funding an AI project. It also helps teams compare how the same work is done across people.
How quickly can we see results?
Bring one workflow. In as little as a week, we map the evidence with you and show what to automate first.
What does Learning capture?
Approved accessibility and interaction signals on macOS and Windows, only while Learning is on. It never captures screenshots, video, images, or audio. Sensitive text is redacted before it is stored.
Where is our data kept?
The raw record stays on each computer and is not uploaded. While Learning is on, a Daily Recap, a separate activity log, is shared with your organization a few times a day. Shared notes and team reports live in your organization's workspace. The Privacy Policy has the details.
Does it deploy automations on its own?
No. Pacific Machines proposes opportunities with the evidence behind them. A person reviews that evidence and decides what moves forward.
Which computers does it run on?
Apple silicon Macs on macOS 11 or later, and x64 Windows 10 or 11. Intel Macs are not supported. Installers are shared after a call.
How can our security team review it?
Start with the Trust Center and the Security whitepaper. Our SOC 2 audit is in progress with Vanta. We are not certified yet.
Work with us
Start with one workflow.
In as little as a week, we map the evidence with you and show what to automate first.