HomeCompaniesAutoComputer

AutoComputer

Desktop RPA with AI computer use

AutoComputer is a desktop robotic process automation system.  Given just a text prompt, our AI automates tedious tasks such as financial data entry by performing all the clicks and keystrokes for you.
AutoComputer
Founded:2024
Team Size:2
Location:San Francisco
Group Partner:Diana Hu

Active Founders

Madhav Singhal, Founder

I am building AutoComputer, the easiest way to automate workflows on your computer without any configuration. Before this, I worked at Replit and was the first engineer on its AI team where I led research and training for all of Replit’s SOTA code models, and worked on data infrastructure, Replit’s core AI services and Replit Agent.
Madhav Singhal
Madhav Singhal
AutoComputer

Rick Sugden, Founder

Co-founder and COO at AutoComputer. Was previously doing my PhD in Medical Biophysics at the University of Toronto where I built the first at-home brain tumour monitoring system with a BCI device. I have spent the last five years implementing and training deep learning networks for biomedical applications including papers published that use ML/DL for diagnostics.
Rick Sugden
Rick Sugden
AutoComputer

Company Launches

AutoComputer helps you automate and accelerate workflows on your computer without any explicit configuration, while keeping you in-charge.

With just a prompt and your recent activity, our desktop AI system starts predicting the next sequence of actions – keystrokes, text input, mouse clicks – that you can keep accepting with a single keystroke. This saves you time and effort while providing you complete visibility and control over the completed tasks.

We’re starting with helping businesses and enterprises automate repetitive, but business-critical workflows such as financial data entry, record keeping, and transaction matching.

Watch our demo video below to see AutoComputer in action!

https://youtu.be/tcFpwrvB7FE


\

The Context

As engineers, we have tools like Github Copilot, Cursor and software engineering agents that help us accelerate and automate large swaths of our programming workflows with AI.

However, outside programming, this is not the case. Many professionals in roles such as accounting, consulting and bizops have repetitive, critical workflows that they want to accelerate or automate with software, but they don’t yet have an ideal solution for this.


\

The Problem

Based on talking to these users, we learnt that the current generation of RPA and AI Agent-driven automation solutions for their workflows have 3 core issues:

Trust and Accountability

For many tasks they need to ensure accuracy and take accountability. So RPA and Agents that just go off do the task for them are not a viable option. They need full visibility and control while the automation executes.

Configuration Complexity

Real-world workflows are complex and span multiple apps and files. Many solutions can’t handle them without the heavy setup, onboarding, or migration required with the current generation of RPA tools or AI agent-driven automation solutions.

Automation Flexibility

Workflows change frequently across formats and tools and they want automations that can adapt on the fly without reconfiguration. Most automations fail on the slightest variation in data formats, schemas or sources introduced when they had new clients, cases and customers.


\

The Solution

With just a prompt, our desktop AI system automates steps for any complex computer usage workflow that they can accept with a single keystroke. It makes repetitive but critical workflows such as financial data entry 2-5x faster, while providing complete visibility, control, and adaptability—all without requiring any setup, recording, or onboarding to a new platform.

Here’s a how it works:

  1. The user opens the AutoComputer app on Windows/MacOS and provides a prompt for their task.
  2. AutoComputer understands the context of the task and comes up with a sequence of automated steps of mouse and keyboard input to complete the task.
  3. The user can simply evaluate and proceed through accepting the automated steps with a single key press, or correct or reject at the granular level without having to fix or re-run the whole workflow.


\

The Team

We are Madhav and Rick.

We met at university ~7 years ago while doing research in Physics. We have been friends and roommates and have hacked on multiple research and side projects together.

Rick recently dropped out of his PhD where he worked on training transformers on brain-wave data for diagnosis, and built the first at-home brain-tumour monitoring system which he patented.

Madhav worked at Replit as one of the first engineers on the AI team where he led research and training for all of Replit’s models, and worked on data infrastructure, Replit’s core AI services and, most recently, Replit Agent.

Based on our experience in AI research and building products like the Replit Agent, we believe that SOTA models are just crossing the threshold of capability where they can start usefully automating and accelerating large parts of knowledge work beyond just software engineering. We are building AutoComputer to realize this for ourselves and for our users.


\

Our Ask

If you are an enterprise or a business with teams doing repetitive back office workflows like data entry, who can benefit from AutoComputer, reach out to us at: founders@autocomputer.ai

If you are an individual or employee who wants to use AutoComputer to accelerate your workflows, you can sign up for our waitlist on https://www.autocomputer.ai/.

Company Photo

Company Photo