By: Jay Feldman

What if you could share your screen with an AI and it could learn to handle the parts of your job you find least appealing just by observing? Imagine screensharing with your new software and having it gradually take over some of your more burdensome tasks, step by step. Picture all that work you don’t enjoy gradually being taken off your plate.

You might think such a scenario is unlikely. However, this is the concept behind the AI tool that Mohammed Nasir, co-founder and CEO of the innovative technology company General Agency, is working on. Her name is Tessa.

Tessa springs from a simple yet relatable impulse: there were aspects of Nasir’s job that he wanted to avoid.

The Birth of a New AI That Personalizes Productivity

“At my previous job as a software engineer at Nvidia, I spent a significant amount of time with administrative overhead and tasks that were necessary but often repetitive — there’s a lot of communication overhead in large organizations,” Nasir explains. “I was able to automate some of it with Linux shell scripts, but the majority of the admin work I had to do involved using a browser, logging in, clicking through various buttons, and doing different things depending on what I saw.”

While training another person to do these activities might be relatively straightforward, Nasir realized that programming AI to handle them would be challenging. At least, that was the case with the technology available at the time. But this began to change last summer.

“Vision-Language Models (VLMs) became advanced enough to navigate websites autonomously,” he recalls. “There were actually a few papers from the year prior, but those were mostly academic and not yet practical enough for production use.”

Nasir recognized that if a browser agent could navigate the web on its own, it could also potentially learn new processes through screen sharing. “I was motivated by my desire to avoid the repetitive parts of my job,” he says. “If I could automate these mundane tasks, I’d have more time to focus on the parts of my work that excite me — like designing systems, writing code, and pushing the boundaries of innovation, rather than dealing with paperwork.”

While Nasir’s motivation may seem unexpected, his background offers some surprising context.

Nasir’s Unique Path to Generating Tessa

Nasir majored in Aerospace Engineering at MIT, not Computer Science.

“The reason I chose Aerospace Engineering was because of the systems thinking and design that the program emphasized,” he explains. “The Computer Science department does have a few courses on systems design, but in Aerospace Engineering, it’s integrated throughout the entire major. Of course, I did research in control systems and related areas on my own, and learning programming was fairly straightforward given the available resources. But systems engineering knowledge is still surprisingly rare, so it made sense for me to pursue it formally.”

Nasir’s expertise in systems is directly applicable to his work in AI development.

“Now, everything I do is about building systems,” he says with a smile. “From systems for delegating tasks, to our internal operations, to the systems that power our product. It all comes back to systems engineering.”

Similarly, his previous role at NVIDIA also involved systems thinking, positioning him to launch his own company.

“I worked on the automotive team, specifically in Planning and Control for self-driving cars,” Nasir recalls. “A lot of my work there involved developing systems that optimized and tuned vehicle controllers.”

In the process, Nasir filed five patents.

“Self-driving cars are, in a way, the first production-grade AI agents,” he says. “Their perception is sophisticated; they understand the world around them. They have a planning layer that determines actions, followed by a control and actuation layer where those actions are executed. There are many similarities between self-driving cars and AI agents for the web. My work on self-optimizing loops for controllers definitely influenced the self-learning mechanisms we use in our current system.”

Thanks to Nasir and his team’s expertise in systems design and advanced technologies, General Agency has made strides toward creating a next-generation AI assistant that could set a new standard in the industry.

Not Your Average AI Assistant

“A lot of companies use the term ‘AI employee,’ but they’re often referring to something no more advanced than an Excel macro,” Nasir points out. “Unfortunately, these subpar products often leave users disappointed, making them skeptical when another ‘AI employee’ appears on their radar — even if this one is more capable.”

This gap in expectations led Nasir and his team to focus on building a product with real substance. “We focus on creating something with real value and let the results speak for themselves,” he says.

To create an AI that learns from screen sharing in real-time, the team at General Agency worked backward.

“We didn’t start with what the technology could do — we started with understanding what people actually need,” Nasir explains. “By examining where their data lives in their workspace and identifying the problems they face, we were able to work backward to find a solution. Ultimately, end users care more about solving their problems than about the underlying technology.”

Capture Your Screen, Guide the Future

Tessa isn’t just a chatbot or another iteration of OpenAI ChatGPT or Google AI Gemini. Mohammed Nasir and his team at General Agency are reshaping how humans interact with intelligent machines, building an AI tool that prioritizes real utility over the hype.

“AI shouldn’t replace the creativity or human qualities that define us,” Nasir emphasizes. “Instead, it should help alleviate the repetitive and uninspiring tasks that people prefer to avoid. That’s why we founded General Agency — to create AI that truly adds value.”