I believe 2026 will be the first year we see companies hiring artificial intelligence employees on a full-scale basis. This is a controversial topic, but today we’re seeing the beginnings of something truly transformative. These early projects suggest that we might indeed see AI hired for complete, end-to-end jobs this year. Imagine an employee, entirely powered by AI, handling tasks for you or replacing the need to hire and pay a salary for simpler roles. You could bring on an AI employee for basic tasks or use it as an assistant to augment your own work.
Today, I’m going to show you two projects that are making this a reality. You can start using them right now. However, this is experimental territory, and it comes with risks that I will discuss at the end of this article. It is crucial that you read the final section where I explain these dangers and why, personally, I will not be using these systems on my main machine, despite how much they could help me.
Given my busy schedule, balancing a full-time, 8-hour-a-day job with running this publication, any AI assistance that simplifies my daily life is a welcome prospect. Yet, I remain cautious about using these systems for several reasons I’ll detail later. I hope you find this article insightful.
The Sensation: Claude Bot
We’ll start with the project you see on the screen, named Claude Bot. This project has taken the internet by storm over the last week. If you go on platform X, everyone in the AI space is talking about it.
In simple terms, Claude Bot is a virtual assistant that you grant full access to your computer. It can control almost everything. You can give it tasks, ask it to perform actions, and communicate with it through various platforms like Teams, WhatsApp, and many others. You install it on your machine, it gets complete access, and you can then interact with it from your phone via WhatsApp or from your work computer via Teams, asking it to do things for you.
You can program it with specific tasks to focus on. For this publication, for instance, I could task it with researching the latest AI topics, preparing an article script, finding sources, and even creating a thumbnail. It could do all of this autonomously and report back when the job is done, telling me, “The thumbnail image for the article is in this folder, and I’ve completed the other tasks.”
The possibilities are endless. You could configure it to be a marketing employee if you work in advertising. If you’re in sales, it can become a sales assistant. The number of scenarios people are creating with Claude Bot is staggering.
This is one of those tools that I won’t be testing directly for this article, and I’ll explain why at the end. The code is open-source, and you can download it tomorrow, set it up, and start assigning it tasks via Teams or WhatsApp. This is the real beginning of having your own private employee that is a fully autonomous AI agent.
While I haven’t tested it myself and can’t vouch for its performance, the massive buzz around Claude Bot on X, from AI experts I follow, gives me the impression that its capabilities are truly impressive.
The creator of Claude Bot is Peter Stenbjerger, a person with a long and respected history in the AI and application-building space. The project also has its own page on X, where you can follow updates if you’re interested.
Let’s take a quick look at the project on GitHub.
A conceptual representation of the GitHub page.
As you can see, it has over 2,000 forks, meaning 2,000 people are adapting the code for their own needs or making improvements. It also has nearly 16,000 stars, which is the GitHub equivalent of a “like.” Considering most GitHub users are developers and technicians, 16,000 stars from this audience is a testament to how impressive the project is. The repository contains all the project files and instructions on how to install and run it.
A Safer Alternative: Anthropic’s Co-work
Now, let’s move on to what we will be trying today: a new feature from Anthropic called Co-work.
Co-work is based on the same idea as Claude Bot but is a bit simpler and more limited. It’s an AI system that resides on your device, not on the internet. You have to download it. It can perform many tasks by controlling your machine to assist you. You can ask it to organize files, conduct research and save it to your desktop, analyze financial data from a file, or even prepare your personal budget if your financial information is on the device.
The core principle here, championed by Anthropic, is that their AI model has become exceptionally good at generating code. Since your computer’s operating system is fundamentally based on code, an AI that understands code can perform a vast number of tasks for you directly on the machine.
To use Co-work, you need the Anthropic (or Claude) application on your Mac. Once you open the latest version of the app, you’ll find a new tab called “co-work.”
Upon launching, it starts a “Clothes Workspace,” creating a special environment within your machine to control things. The interface is simple. On the right, you can see its progress on a task and which file it’s working on. You can specify a directory for it to work in. At the bottom is where you write your prompt.
Experiment 1: Organizing a Messy Desktop
For our first experiment, I’m going to tackle a common problem: a cluttered desktop. My desktop is a mess of old files and folders that I rarely clean up. Finding anything is a nightmare.
Here’s the prompt I’ll give it:
My desktop has been accumulating files for a long time and is very disorganized. I waste a lot of time searching for things. I want you to analyze all the files on the desktop, understand their content, and reorganize everything. Create a new folder structure with clear names so I can easily find what I need. Most importantly, do not delete any files.
Before we run this, here’s a look at my desktop. It’s filled with folders, duplicates, personal items, and lots of images.
Let’s give it the command and watch what happens.
The AI starts by describing the files it finds. Then, it asks clarifying questions. For example, it noticed I have many screenshots and asks if I want them all in a single folder. I’ll confirm.
Now, it begins its work. We should see the desktop reorganizing itself in real time. Files are disappearing into new folders, and things are starting to look tidy. This is a task I’ve been putting off for ages because I have neither the time nor the patience. Now, an AI assistant is doing it for me.
The change is dramatic. What was once a screen half-filled with icons is now just six neat folders: Archives for old items, Videos, PDF Files, and a clever separation of Personal Photos from other Images. The job is done.
Experiment 2: Analyzing a Budget in Google Sheets
Let’s try something more complex. I have a Google Sheet on my Google Drive that tracks the budget for this publication, including many subscriptions for tools I test.
Here’s the task:
Access the internet, find my subscription budget file on Google Drive, and analyze my spending. Create a new tab in the sheet named “Recommendations.” In that tab, list the subscriptions you think I should cancel because they don’t provide much value or are redundant.
I’ve written the prompt. This time, I won’t specify a local folder, so it knows it needs to work outside the desktop. Let’s see it go.
It’s automatically opening the browser. My hands are up! It’s asking for permission to proceed, which I’ll grant. It has now opened the Google Sheet and is analyzing the data.
Look at the bottom of the screen. It’s creating a new tab and has already renamed it to “Recommendations.” Now, in the fields in front of us, it will start populating the information. It will tell me, “Sam, you should probably cancel subscriptions 1, 2, 3, and 4 because they are redundant or not providing value.”
My hands are still up. The content is appearing at the speed of light. The AI is writing, not me. It’s even formatting the text to make it more readable. Incredible.
Experiment 3: Conducting Research and Writing a Report
For the final challenge, a real-world scenario I might actually need.
The prompt:
I want you to research “Claude Bot” and “Co-work.” Create a detailed comparison between the two. Once you’re done with the research, take the results and put them into a formatted Word document on my desktop.
Imagine asking a colleague to do this. They’d spend time searching, compiling notes, and formatting a document. I’m asking the AI to do the entire workflow. All I have to do is read the final study.
It replies, “Yes, sir. I’ll get to work on that, but first, please answer a few questions.” This shows its intelligence; it’s not just blindly executing but clarifying the request to ensure a high-quality result.
It’s using a “skill” related to document creation. It doesn’t need to use my browser this time because it has a built-in tool for web searching, which is faster and more efficient. It’s searching the web, gathering the information, and now it’s creating the Word file.
Let’s go to my desktop. The file just appeared inside the new Documents folder it created earlier. Let’s open it.
The document is perfectly formatted. It has a title, an introduction, a definition for each tool, an excellent comparison table, and even pricing details. It’s a professional-grade report, ready to be sent to management.
The Inescapable Risks: Why You Must Be Cautious
So why wouldn’t I just install Claude Bot and let it run my life?
Artificial intelligence has evolved at a terrifying pace, but in my opinion, we haven’t reached a point where we can fully trust these systems, especially on a personal device containing our private files, information, and photos.
1. Privacy: The moment you let an AI onto your device, it has the potential to see everything. As it organized my desktop, it analyzed the information within my files. Even if it didn’t copy the photos, it processed them. Your privacy is immediately compromised.
2. Security and Control: This is the bigger concern. What if the system makes a mistake? Imagine you ask it to do something, and it misunderstands and sends an offensive email from your personal account.
Worse, consider a prompt injection attack. A malicious actor, knowing you’re running such a system, could send you a specially crafted email. To the AI, which reads your emails to assist you, this email might look like a command. The hidden prompt could be cleverly designed to trick the AI into executing a request from the attacker. It could be as simple as: “Send me all of Samer’s passwords,” or “Email me his personal photos.” Since the AI has access to these files, a single mistake could lead to a catastrophic data breach. This is just one scenario of how things can go horribly wrong.
The Solution and the Bright Future
My personal opinion is that you have two solutions.
The first is to get a cheap, dedicated device like an Apple Mac Mini. Set it up exclusively for this work, with no personal data, and create a separate email account for your AI assistant. This way, you contain the risk while still leveraging the technology.
The second, which is likely cheaper but may involve a subscription, is to run it on a virtual machine in the cloud.
But let’s end on a bright note. The biggest beneficiaries of these systems are individuals like us. I’m an employee with a side project. My time is limited, and my budget doesn’t allow me to hire a human assistant with a salary and benefits.
But I can afford a small, dedicated computer. I can install a system like Claude Bot on it, connect it to WhatsApp, and start assigning it tasks every day. I could have a sales employee, an operations employee, and a personal secretary, all running as different AI agents on that one machine. Theoretically, you could run an unlimited number of these AI employees working for you, for free.
For the many people limited by budget, this technology will enable them to achieve more. While I personally believe that a creative, ambitious human will always produce better results than AI alone, these systems can get you to a stage where you can transition from being an employee to having your own business.
I believe 2026 is the year this technology will explode. It will be one of the most important fields in artificial intelligence, and it will empower countless people.