Ever since launching my first boutique consulting agency/studio, Ruben Arevalo AI & Software Studio, in March 2026 at the age of 24, the first question that came to the top of my head was:
Are we giving AI too much of our agency?
Agentic AI is a concept that's been in the making for decades. We now apply it to every aspect of our lives; answering phone calls, streamlining workflows, and improving HR operations that would normally take hours of work.
However, I've noticed that ever since the launch of OpenAI's ChatGPT in November 2022, people have started to trust AI to make decisions for them. These decisions range from what they're having for dinner, date night ideas, and for people working in the tech industry, shipping production software to accomplish ambitious goals and projects within tight deadlines.
AI has evolved over the past few decades, and has been successful in many use cases, such as diagnosing cancers that normal MRI scans wouldn't have caught and answering difficult questions to problems that would have taken years to solve. It has also extended to retrieval augmented generation (RAG), relying on an external knowledge base that it has never been trained on, thus erasing the knowledge cutoff barrier.
AI, however, has its dark side. Multiple cases of its misuse can be attributed to several instances, such as:
Non-consensual images of people and celebrities being made, depicting them in compromising positions, with the most infamous instance being Taylor Swift back in January 2024 according to an article published by the Associated Press.
Voice cloning scams that have caused both legal and emotional damage, with an instance dating back to July 2025 where a Florida mother handed over $15,000 after receiving a call from an entity cloning her daughter's voice.
A 180% increase in fraud crimes and attempts ranging from falsifying official documents to bypassing stringent automated and manual check systems that would normally catch unauthorized attempts at accessing extremely sensitive customer and financial information held by financial and fintech organizations.
Despite the horror stories that I encountered over the years as AI continued to evolve, I decided to start on a project for my AI studio, which will not only help customers with their questions and automatically submit leads for them, but also implement the same safeguards that will reduce the attempts at fraud while also encouraging users to make their own choices when describing a project and providing their requirements.
I am currently working on a new version of my chatbot called Benny, named after a nickname a coworker had given me to avoid confusion between the multiple Rubens that he has encountered in his life. For the tech stack, I am planning to use the following:
- Python
- FastAPI
- OpenAI API (or Anthropic's SDK library) to handle the automation and submission process of requests coming from my site's chatbot
Benny will be built with the purpose of automating and submitting information to my leads in a structured format so I can fully understand what my client is looking for. The AI agent will be handling the submission of requests, while those same requests are manually and carefully reviewed by me to ensure that they are fully compliant with regulations and state law. This helps create the separation of concerns that aligns with not just software engineering but also creates the line of what can be automated vs manually reviewed by a human/live agent.
Beyond lead qualification, Benny will also serve as an internal HR monitoring tool within J.A.L.E, my studio's business operations + CMS platform. More details about Benny will be published on rubenarevalo.com in the middle of the month.
This brings me to the point I want to share for anyone who is reading this article. When we build something that is meant to improve our lives, we often forget that we are delegating not just our tasks, but also a piece of our autonomy, to an entity that does not have the same nuances and decision-making processes of a human being. There is always a price that we have to pay when building something truly amazing that can be both very beneficial but also maliciously exploited.
For example, if you're a hiring manager, and you're looking to hire a commercial truck driver using solely AI, it will definitely make things easier for you and allow you to focus on the critical aspects of the process, such as the application review, in-person screening interview, and verbal examination, though the process varies by company.
However, there is one critical part that may be missed during this process, and it is the background check. Because the employer decided to use only AI for the interview process, they forgot to check that the driver they hired has a history of driving under the influence, on top of having their license suspended twice. Again, this is only an example, but it highlights the discrepancies that AI or its agentic counterparts may cause if not balanced carefully with our own judgment, which is a crucial part of our agency as human beings.
A second notable example would be the use of AI by students. While AI has helped students grasp concepts better, it has also impacted our critical thinking abilities as revealed in findings of a study released by MIT back in November 2025. AI has helped us accomplish very ambitious projects and goals ever since OpenAI's ChatGPT was introduced in November 2022, but it has also cost us pieces of our own autonomy. By depending more on AI, the neural pathways in our brain that map to and are critical to our creativity, critical thinking, and decision-making skills are slowly reduced depending on how dependent we become on it. This same dependency especially impacts children, teenagers, and young adults under the age of 25, the age at which our brains reach full development.
Lastly, I am currently working on a separate project as part of my studio's goals. The name of the project is J.A.L.E, which stands for Job Automation & Logistics Engine. J.A.L.E. is based on the Spanish word for "jale", a slang word that is used in Latin America to refer to a job, work, or occupation. J.A.L.E. takes CMS and operational management concepts and merges them into one system, ensuring that anyone who utilizes my product in the future not only connects it to their website, but is also used for tracking project management, which will be reflected in a statement of work (SOW), capturing the entire scope of the project and what was done while also noting down the individual parties who were responsible for carrying out certain aspects of a project.
But the most significant role agentic AI plays in J.A.L.E isn't in the statement of work, but with payroll, a very common issue small business owners face. In addition to Benny's capabilities of answering and submitting inquiries from potential leads in my website, it will also have instructions to review discrepancies in employees' timesheets, such as detecting whether employees have worked a certain amount of hours (e.g. 40 hours in 3 days) during the week, if they have not clocked out, if their pay does not match the hours they actually worked, and more. These discrepancies will be flagged for manual review by a human who is specialized in HR management and operations to ensure fairness and consistency across the board.
By implementing this approach, it further supports and adds to the notion that a separation of concerns is necessary when dealing with an entity that is nuanced but is prone to making errors, and another entity who is less likely to make them but does not have the nuances and understanding of human nature.
As someone who is building their own AI studio and agentic AI, I believe that somewhere along the way, there were moments where we gave away a piece of our autonomy to AI when we look for product recommendations, recipe ideas, and more. Even though we only gave a tiny piece of our autonomy and trust away to an entity that lacks human nuance, that's all it takes. And unfortunately, if we continue this pattern without balancing our own judgment and needs, we will eventually forget what made us who we are in the first place. It will reduce the hard work we have put in ourselves along the way into a mere shadow of ourselves.
Now the important question we must ask ourselves, where do we draw the line between what we decide and what AI can decide for us, both professionally and personally?
Top comments (0)