Software Development in the AI Era

What's the local IT community's take?

Slovácké IT pivo is a meetup held in Uherské Hradiště where IT professionals from across the Slovácko region of Czechia meet. On the 12^th meeting, I was asked to moderate a discussion on the topic of Cloud and AI. This gave me an opportunity to ask these colleagues the following question:

Mark Zuckerberg says that this year artificial intelligence agents will start to be employed as mid-level engineers. How will AI impact your job?

Though answers were given in sentences, I divided them into three categories based on how I understood them. The distribution is shown in the following chart.

Around 50 people attended the event, and 21 of them shared their views on the topic. Fewer than a third expect little to no impact. Here are some of their responses:

    
• “More likely as an auxiliary tool. I do not expect any major changes”
    
• “Seeing how Copilot and other chats respond, I have no fear" 😊”
    
• “The most it can do is help me generate simple code” 

Out of the answers, that acknowledge an impact, albeit with unclear consequences I would like to highlight the following: 

    
• “I will spend less time explaining things to juniors and more to AI.”
    
• “We will run out of juniors”
    
• “I think we will need less programmers. The more senior, the later we will be replaced. There will be new roles, but more will disappear.” 
    
• “AI = another junior on the job.”

Concerns about the future of junior engineer roles seem frequent. However, changing to a more positive note, almost half of the respondents expect positive impact:

    
• “It will make me much faster at my job “
    
• “It will make it easier for me to enter new technology “ 
    
• “It will be harder to start off in IT and other areas. Use of AI will be necessary to reach maximal productivity.“ 
    
• “It can help, but I am not afraid of losing my job.”
    
• “It will make coding more effective. Remove discussions with difficult colleagues. It will shorten the time to build MVP.” 

Large Language Models, Chatbots and Edhouse

It is worth pointing out that the term AI in the results of my small survey actually refers to a specific area of AI: chatbots powered by large language models. This section will focus on this type of technology as it seems to be currently influencing the work life of developers the most.

Among the main uses of these tools are conversations with an assistant through a web interface (ChatGPT, Gemini, Claude) or integration of an assistant into a development environment, for example the embedding of GitHub Copilot in Visual Studio Code. The Cursor tool deserves a mention as it is becoming increasingly popular.

In Edhouse, we conducted another small poll in this area. We asked our colleagues the following:

“How much time does GitHub Copilot, or a similar tool save you during programming?“

This is how 19 participants responded: 

To save you time crunching numbers, I calculated the average time saved by Copilot or a similar tool to be almost 15%. We cannot work out the total increase in productivity when solving problems for our clients, however given programming makes up the most of what we do, we assume it is significant.

This technology is not without risks. We have an important obligation to protect our customers’ data and ensure that we always operate in agreement with our NDAs. Therefore, we established guidelines for AI use in our company and a process for the approval of new tools of this kind. We tightly cooperate with our customers in this area to ensure the safety of the data entrusted. An internal conference held with onsemi titled AI copilots, their characteristics and practical applications is one example of such an experience exchange.

Author’s closing remarks

Given that I spend a large part of my work time doing analysis, I do not use these technologies just for programming. I often need to analyze data in Python using Jupyter Notebook where AI helps me generate code. One of the advantages is being able to describe how I would like to process the data and what patterns to highlight using natural language. I of course have to verify all the methods applied, but this saves me a lot of time that I otherwise would have spent reading documentation.

Data analysis would be pointless without presenting its findings to anybody, which involves a lot of creative writing. You do not need to fear, this post is not AI generated. AI just helps me look up synonyms and expressions. Believe it or not, this also saves me a great deal of time.

I believe that the advantages of AI for software development and other creative activities are now evident. We naturally fear that the more we offload to AI, the more we fall behind in these areas. It might come true to some extent, but those already hitting limits to their creativity will be able to achieve remarkable feats.