Building Software That Listens, Thinks, and Acts

For years, how we interact with software, whether we're managing office tasks or work in places like warehouses, has been stuck. We click through menus, try to figure out complex dashboards, type information into boxes, and sometimes struggle with stiff or confusing screens. Or we might spend hours reviewing recorded video trying to find out what happened after a problem occurs. Most of the time, we're just trying to do two basic things: ask questions to get answers or take action to get something done.

But what if software could proactively do more? What if it could understand what you need and handle tasks for you, working across different systems smoothly? Imagine AI agents navigating complex systems for you. By AI agents, I mean smart software programs designed to understand requests, make decisions, and perform tasks on their own, much like a helpful human assistant might. Picture asking, "Show me any damaged shipments from Dock 3 this morning," and instantly getting the right video clips and a quick report. Or requesting, "Reset my VPN password and notify IT," and having the whole thing handled automatically. This isn't some far-off dream. A new way of using software, sometimes called "Service as Software," is starting now. It means smart AI agents that can understand text, voice, video, and actually do things can make work easier and help us get more done. At Cortexa Systems, I'm working hard to build this future.

Today’s tools often ask too much of us. An IT manager trying to fix a helpdesk ticket might spend hours jumping between different tools, reading logs, and passing the issue along, while their coworker just waits. A warehouse supervisor might have to manually review hours of camera footage to figure out why safety incidents keep happening. A finance analyst needing a report might have to search through different systems and wait days for approval. An HR person trying to send out new hire paperwork might get stuck because of confusing workflows. These kinds of frustrating roadblocks happen in digital office work and in physical operations like warehouses. Both worlds suffer from similar system-wide problems that make work harder for people.

At Cortexa Systems, I believe software must adapt to people and the reality of their work. The AI agents I'm building are designed to fix these problems. I'm inspired by how the human brain works, particularly the parts responsible for thinking (neocortex) and memory (hippocampus). Like the brain, Cortexa’s AI agents pull together different pieces of information, think through problems, and learn over time. Just as our memory connects new experiences to what we already know, the AI agents I'm building link different data points to provide useful insights and take smart actions. This idea from biology pushes me to build AI that acts more like a smart partner than just a simple tool.

A simple request like, "Reset my password and give me access to the Q3 folder," could trigger an AI agent to check who you are, make the changes in the necessary systems, and let the right people know it's done. Similarly, an instruction like, "Flag all instances of improper pallet stacking on outbound shipments today and generate coaching alerts for the relevant supervisors" allows an AI agent to analyze visual data, identify the issue, link proof, and start a process to help fix it, turning stored data into helpful action.

Using different ways to communicate and understand, like text, voice, and analyzing video, is important. Cortexa’s AI agents can understand more context this way. In IT support, an AI agent might text diagnostics and use video to guide a fix. In HR, it confirms start dates and syncs payroll. In Finance, it approves expenses or flags unusual numbers. In the warehouse (FloorSight AI), it analyzes video feeds to detect safety hazards like near-misses, checks if protective gear is being worn, spots damaged goods or poor handling, and connects what it sees on camera with information from the warehouse system to give amazing insights into what's really happening. These abilities help people out, removing barriers and letting them focus on more important work.

While the big AI models you hear about (Large Language Models, or LLMs) are powerful, sometimes you need something more focused. That's where smaller AI models (sometimes called SLMs) built just for specific jobs come in. Think of AI trained specifically on internal company policies or on recognizing warehouse safety rules. They can be more accurate, faster, and better for privacy because they're trained on specific information safely. At Cortexa, I use the right AI tool for the job and make sure they keep learning within each customer's setup.

Security and privacy have to come first. As someone with a background in security and privacy engineering, I know how critical this is. Cortexa’s AI agents are built with privacy in mind from the start. They often process data right within the customer’s own systems, keeping information safe and meeting strict rules. The tools I'm building help companies keep control of their information and follow the rules, which is vital whether handling employee information or sensitive operational videos.

This isn't just an idea for me. It comes from seeing problems up close. Last year, helping a hospital deal with a huge backlog of helpdesk tickets, I saw how broken digital support slowed everything down and frustrated people. Employees waited weeks for simple things; support teams were drowning. Around the same time, wanting to understand work beyond a computer screen, I took jobs working in warehouses. I drove pallet jacks on night shifts, counted inventory, loaded trucks. I saw a similar struggle there: tough physical work made harder by process problems. Picking mistakes cost money. Badly stacked pallets created problems for truck drivers. Safety issues were often dealt with only after an accident, by digging through old video. Companies used cameras to monitor drivers or track exosuits mostly to lower their own risks, but that same focus wasn't always there to proactively help workers do their jobs better or safer day-to-day. I saw the gap between all the video being recorded and the lack of useful help coming from it.

Seeing these similar patterns of frustrating problems slowing people down in both office support jobs and warehouse jobs was my turning point. It showed me this isn't just about one specific issue; it's a widespread problem with how our tools support us (or fail to). It became my mission to build something better.

So, in March 2025, through Cortexa Systems, I launched WorkSync AI, my AI assistant for internal support (IT, HR, Finance). At the same time, I began developing FloorSight AI, applying the same ideas about AI agents to the physical world of warehouse operations. It uses AI vision and intelligent agents to tackle quality, safety, efficiency, and rule-following challenges I saw firsthand and that industry reports confirm are major issues.

These products come from real frustration and my belief that there's a better way. We need systems that understand, act, and improve. But building them means being honest. As I wrote recently, the AI world is full of buzzwords, and the tests used to measure AI aren't always reliable. Models trained on test answers aren't really smart; systems secretly run by humans aren't really automated. That's why my approach at Cortexa is grounded in reality. I plan for mistakes: if an AI agent gets stuck, it should admit it and ask a person for help. I test the AI on real job tasks: does it reliably reset passwords, or accurately flag damaged goods in a real warehouse? That's what matters, not just some generic test score. I set honest expectations: AI isn't perfect; it makes mistakes. I aim for steady improvement and being upfront about limits. And I prioritize trust: making sure data is secure, privacy is protected, and the impact on workers is considered is key from the start.

Having worked these jobs, I know the physical toll and the risks involved. My goal isn't just to automate tasks away; it's often to automate the dangerous, the tedious, and the frustrating parts, making the existing jobs safer and more effective. I believe technology serves us best when it assists and elevates human capabilities. While the long-term impact of AI on the economy is something society needs to grapple with, my focus is on building responsible tools today that help people succeed in their current roles and make operations run better and more safely. Understanding how hard it is to use real-world data, especially sensitive video or personal info, means making sure privacy and fair treatment for workers are built-in from day one.

Alongside building products, my consulting work helping companies use new AI tools to solve real problems has grown. This shows me the need is real; companies want to use AI but often need help making it work practically. Seeing big companies like Accenture make billions from AI services shows how big this change is. Experts like Foundation Capital and Marc Benioff are right: the change to Service as Software is huge. It allows AI to tackle operational challenges in areas where trillions are spent globally on labor and services, aiming for better outcomes and efficiency. This shift means AI can handle repetitive, time-consuming, or even dangerous tasks, boosting overall efficiency, improving safety, and making operations more resilient. FloorSight AI brings this potential to warehouse tasks, and WorkSync AI streamlines internal support roles, freeing people up for more complex work.

My focus now covers three areas: Software (like WorkSync and FloorSight, which I build through Cortexa), Consulting (where I help companies use AI practically), and Infrastructure (the basic tech needed to run it all). I'm not just trying to fix old tools; I'm trying to rethink software so it acts like a helpful partner in getting work done. I'm building software that listens, software that sees, software that thinks, and software that truly works.

This is the age of Service as Software. And I am all in.