Today, software development has become a widely used term. But how many of us have wondered how software ends up on our computers?
AI or Us? What could the future look like?
Lately, I've been hearing the question, "Could Artificial Intelligence (AI) replace us?" Honestly, at that time, I didn't have the fear that it could replace us or potentially cost us our jobs. So I started to delve deeper into the subject.
Lately, I've been hearing the question, "Could Artificial Intelligence (AI) replace us?" Honestly, at that time, I didn't have the fear that it could replace us or potentially cost us our jobs. So I started to delve deeper into the subject. Obviously, I came across ChatGPT from OpenAI and thought, "Sure, this is it!" But after some searching, I found other tools as well. So suddenly I had access to dozens of AI variants. Since I found ChatGPT, I set out to see what it does and doesn't do. I wanted to understand how and if it could replace me at work. I'm a Business Analyst (BA) and Proxy Product Owner (PO) at PitechPlus, and to determine if AI could replace me, I first detailed my daily tasks.
What does a Business Analyst do?
In the morning, coffee is a must, then I start working: I solve problems, attend meetings, or write down the requirements received from customers. Usually, my colleagues ask me how I can handle so many tasks and how I can successfully solve them. This is what a typical day looks like for me. It seems simple, but when I describe everything I do throughout the day, I realize how complex it can become. So, a BA is the person who proposes solutions for the business, helps the client get the business up and running, analyzes and understands the client's needs. Then, they convey these details to the development teams in clear and easy-to-understand language.
In the literature, we find mainly two types of tasks: one that is profound and the other that is shallow. Thus, some tasks require creativity and comprehension to be completed (the profound one), while others are repetitive and could be performed by a robot (the shallow ones) - this is where the contribution of artificial intelligence comes into play.
What is Artificial Intelligence?
The concept of AI is quite common today: we see it in the news, encounter it in new applications that boast integrated AI, and even in government initiatives (such as ION). AI is actually a form of machine intelligence that can perceive the environment and surrounding information and optimize it. When we mention ChatGPT we are actually talking about a sea of information in a LLM (Large Language Model), where there is an algorithm that quickly searches through all the available data and returns a response in a format as human-like as possible. But the information doesn't stop here. There are many more AI tools based on different technical solutions and ideologies. t's worth noting that this article doesn't delve into the origins of AI but rather focuses on how to use it.
What do we not use AI for?
Having defined the relationship between Business Analysis (BA) and Artificial Intelligence (AI), the next step is to identify how artificial intelligence can be used in everyday work. To understand what we can use it for, we first investigated what we cannot use it for.
First and foremost, AI is not capable of speaking with a customer (or a colleague), so this part remains our responsibility. However, even if it could understand what a client is conveying, it would not only need to interpret their requirements but also ask a series of questions to clarify many aspects. This is called "prompt engineering," which is a task initiated by us towards AI, not the other way around. Of course, one could argue that AI is capable of reading transcriptions, but after using transcriptions several times in calls, it becomes quite clear that they contain significant errors. Therefore, we cannot rely on transmitting this text to AI because the results would also be erroneous.
Secondly, AI lacks creative thinking skills, inventiveness, or the ability to perceive mimics and emotions. AI will never come up with information it hasn't "read" somewhere because it is not capable of generating new thoughts or ideas. Consequently, we cannot assign tasks that are part of deep work to artificial intelligence.
On the other hand, AI is very capable of performing shallow, repetitive work tasks.
Last but not least, AI rarely provides the source of the information (obviously there are exceptions, but here we are talking about the most common ones like ChatGPT) - basically, it will not indicate whether the given text is plagiarized or to what extent it is plagiarized and, most importantly, it will NOT necessarily provide a correct answer to a question.
What do we use AI for?
Now that we've discussed what we can't do with AI, it's time to approach it from the angle of how we can use it.
This is where the concept of shallow work comes in: shallow work is often repetitive, lacks novelty, can be tiring, and, above all, time-consuming. For example: writing user stories, user templates, acceptance criteria, and fast learning. Let's see how we can optimize these tasks using AI.
First and foremost, an extremely useful application of artificial intelligence is related to fast-paced information retrieval. When we are in a call, analyzing a potential new project, or even a new idea from a client, and we come across a concept about which we have no knowledge, it is very productive to ask an artificial intelligence, thus getting the answer in a matter of seconds. Of course, we could ask a colleague, but receiving the answer could take some time. When we are in an online meeting, we can research the topic on the spot without having to wait.
Secondly, we have observed that AI tools excel at constructing "user templates," "user stories," and/or "acceptance criteria." However, when compared to rapid learning, here a BA needs to intervene much more: in learning, the BA must decide whether the information provided by AI is relevant and true or not - this is done using critical thinking skills and past experiences. On the other hand, when we want to build a "user template," it's not enough to ask AI what it looks like or to request it to construct one. Rather, a method called a line of questioning must be employed: we need to have a goal, meaning where we want to reach, know what questions to ask to get to that point, and most importantly, not allow AI to deviate from the subject, while we insist on the path we want to take.
Thirdly, we explored two more technical areas where AI could provide assistance: coding and Excel analysis. Regarding the first topic, we quickly realized that without prior user experience, AI would continue to provide code. However, it often contains errors, mostly because it does not take into account other parameters, and the user does not know how to follow a line of questioning. The same situation arises in the case of Excel analysis. However, with significant past experience, we managed to conduct a more detailed study: on a large Excel file (over 100 columns and over 75,000 rows), we needed to extract comparisons and various statistics for certain groups. Using simple methods, it was not feasible to obtain all the data because it was physically impossible; another method was to write complex formulas that took into account different sheets in Excel, multiple columns, and conditions to result in a comparison. They could be written by hand, but even for someone very experienced, it would have taken time. Thus, we chose to write these formulas with the help of AI: we created the idea of a goal (the intended result), then considered each necessary column and element to reach this goal, and then explained to AI that I needed a formula that would extract data Z from sheet X for user type Y.
AI built these formulas in just a few seconds. Following that, human validation was performed: it was discovered that the formula was incorrect because AI used only a comma instead of a semicolon. After a quick change, also with the help of AI, it was noticed that the final result was in a strange format. The result was analyzed, and it was realized that the formula was not well thought out. Therefore, another change was requested from AI.
Overall, the outcome was that AI constructed the formulas based on well-defined information provided by me, and then corrected them considering my validations. The result: a reduction of approximately half of the total analysis time for the file.
Concluding "AI or Us?": Whose future is it?
Leveraging multiple experiments and tests, we have arrived at a conclusion for the question: "Does AI replace us?" The answer is YES and NO, even though it may sound strange.
From my perspective, based on the daily tasks of a BA and what AI can currently offer (probably in the near to mid-term future as well), artificial intelligence without human intervention is not capable of completely replacing a BA, a Developer, or an Analyst – hence, the answer is NO. On the other hand, a BA who knows how to use at least one AI tool (but preferably more) has a very good chance of replacing a person in the same position who does not possess a similar skill set. A person with well-defined basic knowledge and the ability to use AI technology will be more productive and YES, can replace someone who doesn't possess these skills.
So, for me, it's quite clear what the future will look like. The question "Will AI replace us?" is partially valid because, as long as one does not adapt, a person will be replaced by another with more experience. This is not a novelty, and AI is not an autonomous threat (it's not about "Skynet"), but rather a step forward towards constant evolution.
Read original article here.