The Brave New World of AI and IT: Challenges and Opportunities
“Nothing is as constant as change.”
(Heraclitus of Ephesus, 535–475 BC)
Welcome to the Future
It’s the year 2033. The hype surrounding LLMs and ChatGPT started ten years ago and has since had a major impact on software development. The impact didn’t happen with a big bang, but rather happened gradually in several phases. It all began in 2024 with the use of artificial intelligence in the form of coding assistants.
Phase 1 (2024): Use of Coding Assistants
Figure 1: Phase 1: Use of coding assistants in development teams
So that’s how it started: Every developer in the team could use an AI coding assistant if they wanted to. As the developers were very tech-savvy, they gladly accepted the offer. The application proved to be a success. The developers soon became friends with “their” new assistant and they got better and better through constant learning. Looking back, this was similar to the introduction of navigation computers in road traffic in the 2000s and the introduction of smartphones in the 2010s. Development was significantly accelerated by AI support and management was delighted. However, there was soon a desire to support other areas of development with specialized AI assistants. After two years, IT therefore moved on to phase 2.
Phase 2 (2026): Use of Specialized Assistants for UX, DevOps, Software Testing and Software Architecture
Figure 2: Use of specialized AI assistants for text, architecture, DevOps and UX
The development teams had acquired a taste for it and were also busy using other specialized AI assistants. These were often assistants for cross-cutting topics, such as the design of graphical user interfaces (UX), DevOps, IT security, software testing and software architecture. The direct use of AI assistants eliminated a lot of waiting time for developers, as the cross-cutting issues could now be dealt with directly by the team itself. Conveniently, these tools also took over the documentation. Due to the great success of the first two phases, the management became bold and wanted to further increase the productivity of the teams by giving artificial intelligence more responsibility for the success of the project. Phase 2 was finally followed by phase 3 with AI team members.
Phase 3 (2027): AI Assistants Become Full Members of the Development Team (20% AI)
Figure 3: AI assistants become full members of the development team
AI assistants now became full members (AI developers) of the development team. In addition to human developers with AI assistance, there were now individual AI developers. Initially, they were assigned the tedious tasks that their human colleagues didn’t feel like doing. Gradually, however, the teams realized that tandems of human and AI developers can be particularly productive in solving tricky challenges. And so collaboration between human and AI colleagues was further encouraged. However, new challenges also emerged during this phase, particularly when it came to who should be in charge: Human or Artificial Intelligence? There were also conflicts of interest between the AI assistance systems themselves, which were resolved by the humans in the team in phase 3. However, it was clear that the Agile Manifesto, which was originally created for human teams, also had to be adapted. The result of the transformation was the AI Agile Manifesto, which we will discuss below.
Phase 4 (2028): AI/Developer Tandems Form the Entire Development Team
Figure 4: Development teams consist of AI/developer tandems
After the AI colleagues had proven themselves in solving tricky problems and implementing unpopular tasks, confidence in the developer AI had grown so much in 2028 that the structure of the development teams was fundamentally changed: tandems of humanoid and AI developers were now formed, so that the AI share of the teams grew to 50%. During this phase, the performance of the teams increased significantly, as the AI developers were able to work around the clock, which meant that the 14-day sprints could be shortened to 5‑day sprints. Reviews and tests are largely no longer necessary. The main task of the human developers was to coordinate, monitor the results and resolve conflicting goals. However, the AI colleagues learned diligently and the management and clients were enthusiastic about the high productivity, which led to phase 5, the complete takeover of all development activities by AI colleagues.
Phase 5 (2030): AI Completely Replaces the Entire Development Team (100% AI)
Figure 5: AI completely replaces the human members of the development team
Phase 5 was activated in 2030. This resulted in the complete takeover of all development activities by AI developers. This included development, testing, the design of graphical user interfaces, DevOps activities, first and second level support, IT security and the development and maintenance of the software architecture. By replacing human colleagues, it was possible to further increase output so that sprints no longer took one to three weeks, but just one day. And there was even another evolutionary stage within reach: the direct creation of machine code by AI developers.
Phase 6 (2033): AI Generates Machine Code Directly
By eliminating the human component in the development teams, it was no longer necessary to take all the many intermediate steps from planning to implementation and going live that human colleagues needed in order to understand the problem and the software. In phase 6, the direct path from the customer requirement to implementation in machine code was realized. A quantum leap in software development.
Today, all software development is fully automated and has been accelerated by a factor of 10–100. Today we talk about one-day sprints. Stakeholders now speak directly to specialized chatbots and discuss new requirements and their implementation with them on the same day. This finally fulfills the long-held wish of clients to implement their wishes and requirements without the disruption of IT.
This development was made possible by the consistent expansion of artificial intelligence, the willingness of AI assistance systems to learn and the adaptation of the Agile Manifesto to the “new era”: the AI Agile Manifesto.
The AI Agile Manifesto – Artificial Intelligence
- You need to know what you want to achieve with AI. There is a trade-off between feasibility and business impact.
- The organization must be committed to the AI project.
- The leader of the AI team must be an effective manager and a leader who has a clear vision of AI.
- Design thinking and agile are valuable tools. Focus on the to-do list to control the scope, cost and schedule of the AI project.
- You need to know all the factors that can influence the AI project.
- The AI project must leverage and align with all of the organization’s process resources.
- An AI project needs excellent people, models and data.
- AI quality is not only about the quality of models and software, but also about people and data.
- AI risk management requires constant risk assessment, a risk strategy and human-in-the-loop.
- You need to involve all stakeholders and have a clear communication plan, especially if something goes wrong.
- Recognize, understand and address ethical concerns caused by AI.
- Agile for AI requires a specific approach with longer cycles and more exploration.
Conclusion
We hope this outlook on the next 10 years of software development, taking into account the use of artificial intelligence (AI) without blinkers and prohibitions, has shown you that we are actually on the threshold of a different future today. A future that is very different from what we may have imagined so far – but a future that we are not at the mercy of, but one that we can still actively influence. The next blog article will deal with the use of AI in software architecture work. Why not take a look?
Sources
https://hups.com/blog/are-developers-needed-in-the-age-of-ai
https://hups.com/blog/agile-in-the-age-of-ai
The AI Project Handbook: How to manage a successful artificial intelligence project (The Artificial Intelligence Handbook Series) von Minh Trinh PhD (Author)
https://www.wbscodingschool.com/blog/is-web-development-dead-in-the-age-of-ai/
This is a translation of ITech Progress’ blog post “Die schöne neue Welt von KI und IT: Herausforderungen und Möglichkeiten”. Here you can find the original blog post in German.
