AI that works beside you, not instead of you.

Jacard AI is a deep-tech Human-AI Research Lab that ships.Our research yields Computer-Use Agents that can collaboratively work alongside humans to learn and execute an enterprise team’s workflows, so work gets done faster and with fewer errors.

The problem

AI agents can now converse in natural language, see images, reason, write code, read screens and act on them alone. There is no doubt this will mark a stark shift in how humans work with computers.

Meanwhile, our expertise as people will continue to be jagged: we each have areas of strength and areas where we need help.

Today’s agents could be that help, but they are made for public software and for tasks anyone can describe, and they are better alone than beside someone. Most work inside companies runs on proprietary software and workflows no one wrote down. Often the only record of how it is done is the person doing it. That is where Jacard steps in.

Our core mission is to amplify human potential by giving every person and organization an AI system that truly fits them. One that knows their software, learns the nuances of their workflows, fills the gaps in their expertise, and takes on more of their work a little at a time. Today, we pursue this through computer-use agents.

One principle

This has long been a holy grail of Human-Computer Interaction (HCI) research, and today’s models can get us there. We are building on forty years of findings on how people and software should share a task, which today’s agents do not use.

We differ from other AI labs in one key principle. Other labs build for autonomy. The measure of their agents is what they can do alone, and working with a person is added afterwards. We build for collaboration: AI agents and people working together.

CoCo Agents

We call the agents natively built this way Collaborative Computer-Use Agents, CoCo Agents for short. A CoCo Agent works on your screen while you do, with a pointer of its own beside yours. It natively understands how people work together, knowing when to lead and when to follow, when to speak and when to stay quiet, when to hand work over and when to pick it up. It learns your software from you the way a new colleague would. It takes a few things off your hands, you correct it, and it takes a few more.

Where we are

We are in stealth, and Jacard’s CoCo Agents are being used by and co-developed with enterprise design partners for new user onboarding, software adoption, and collaborative human copilots for expert work.

This is a hard problem, and we reinvent the whole stack for it: the infrastructure, firmware, the models, the environments they learn in, the core tenets they operate with. We believe in open research, and do it through the Human-AI Research Consortium.