AI will work instead of me!

We are covering the effect AI has on professional life and beyond. The first post covered the employer’s perspective, in this one we will look at the employee’s point of view.

People are so hyped with the capabilities they see in AI agents that many now think that they can just relay all their work to an AI and then present its results as their own and claim their salary. Obviously, if you do nothing in the process sooner or later you will be recognized as a redundant broker and removed from the equation.

AI can be used in multiple ways to effectivize your delivery. We will briefly discuss some dos and don’ts we have learned out of practice while using AI agents in our software engineering work.

AI can not help you if you do not know how to validate its produce

This is something that we have seen over and over again. Engineers that act as proxy to an AI agent deliver whatever code was spawned by the agent without ever employing a single thought in the process. However, while the AI code generation can help you speed up your delivery it is almost never the case that you do not need to tweak / fix / improve what was generated to get it fully working. If you do not know how to validate that the piece of code is solving your task you will not be able to make this step and you will fail your assignment. With or without AI, never engage in a task you do not fully understand. This might sound banal, but a really great number of engineers fail this simple rule.

AI is best used to solve specific tasks

AI delivers, with better results for better defined, specific tasks. A great engineer is one that is capable of doing divide and conquer: breaking down a more complex, high level and abstract task into several more concrete and scoped subtasks. This ability has always been a strong quality for a professional of any occupation, but now with AI “workers” it might slowly turn into one of the most pivotal for the best professionals. The analogy that can be made is with playing cards in England in the old days: masters did the calls and then servants did the card play. Finding your way through a complex task and breaking it into smaller tasks requires creativity, ingenuity and is exciting, this is why we do it. On the other hand, the smaller tasks themselves are very often more straightforward and even boring. Let AI be our servant for this.

AI is best at solving commonly found problems

AI is really good at solving problems that many people encounter: implementing an API integration, generating boilerplate, implementing specific algorithms and some parsing and scraping. The more domain specific you go the less efficient the AI facilitation. When you realize this, this is good news for the owner of the business you work for. So many of them are highly sensitive to privacy and IP protection and this matters most on domain specific tasks. Such work is best kept off public analysis / knowledge base.

Learning with AI

Many people have now completely discontinued using google search engine in favor of AI chatbots. This practice holds its disadvantages overall as the AI generated results are not always trustworthy. However, we will focus specifically on one of the possible purposes of this interaction: this is AI-facilitated learning. Our take on this is that AI can be good to help you compose a plan / strategy for learning something new, but one should not use it as a single source of truth. This is again due to the fact that AI summaries are sometimes inaccurate and misleading. The proper approach will be to use AI to suggest sources to learn from. As a second step you should verify these are credible knowledgebases (an imperative step every time you want to learn new information). Finally you have to verify that the AI suggested summary matches the original source’s contents.