AI's Impact on Software Development Teams

Introduction to AI in Software Development

During the end of 2022 and most of 2023, AI experienced exponential growth both in adoption and technological leaps. Suddenly, everyone started to use this new version of AI that can generalize concepts and, most importantly, in plain natural language. This made the technology accessible and approachable to anyone. People are discovering new ways to work, with the collaboration of AI, increasing their productivity or as a virtual expert collaborator on almost any topic.

GitHub Copilot: A Milestone in AI-Enhanced Coding

With the arrival of large language models or LLM, data scientists started to do all kinds of experiments. But what is a LLM? Simply put, a Large Language Model (LLM) is an artificial intelligence (AI) program trained on vast amounts of text to understand existing content and generate original content. LLMs are built on machine learning, specifically, a type of neural network called a transformer. They are notable for their ability to achieve general-purpose language understanding and generation. LLMs can recognize, generate, translate, and/or summarize vast quantities of written human language and textual data. They acquire these abilities by using massive amounts of data to learn billions of parameters during training and consuming large computational resources during their training and operation. In particular, a few years ago, one of these experiments, performed by GitHub engineers, was to ask to Open AI LLM to solve a challenge designed for developers. To the surprise of this team, the model provided code that worked. They discovered that most of their input challenges were solved and coded correctly.

That was the beginning of code generator AI tools, and the most successful is GitHub Copilot. In their latest developers event, Universe 23, they announced that they currently have over 100 million subscriptions between individuals and company users and over 37,000 companies of all sizes. The tool is becoming increasingly powerful, but most important is the exponential velocity of releasing new and larger models and features. The following is a summary of milestones to visualize the increment in the speed to deliver new versions with more complex and powerful features:

• June 29, 2021: GitHub announced GitHub Copilot for technical preview in the Visual Studio Code development environment.
• October 29, 2021: GitHub Copilot was released as a plugin on the JetBrains marketplace
• March 29, 2022: GitHub announced Copilot's availability for the Visual Studio 2022 IDE on
• June 21, 2022: GitHub announced that Copilot was out of technical preview and is available as a subscription-based service for individual developers.
• Feb 14, 2023: Copilot became available for Business.
• July 2023: Copilot Chat was released for Beta. This potent tool is powered by GPT – 4 (GPT powers Copilot code generator – 3 until November 2023).
• November 2023: Copilot announced that in December 2023, they will release the new Copilot based on GPT – 4 and fully integrated with most of the Development Cycle aspects, not only in VS Code and all the IDEs of JetBrains but also in the GitHub.com portal, allowing direct interaction with project files and PRs using GitHub Cloud. Also, the new Enterprise subscription will be available in December.

Finally, the roadmap for 2024 is to release Copilot Workspaces, which basically will analyze a requirement in the whole project, a GitHub issue, for example, provide insights into the current state, a proposal to implement it, a plan will implement it, create the unit tests, run it, test it and deploy it.

The Future of Software Development with AI Tools

Other tools are available in the market. For example, Google announced Codey in May 2023 and created a complete cloud-based development environment. The underlying technology that powers Copilot is Open AI Codex, which is based on their GPT LLM. Hence, with the release of GPT 4, an entire project can now be analyzed, and the same is true for an entire book of 300 pages, something impossible with the previous GPT-3 model. Open AI announced that prices are also going down. Hence, GPT – 4-based tools are also a great source to empower Software Development Teams. A feature that Copilot Enterprise is providing, and that will be a total game changer, is the capacity to gather all the technical documentation of a company, a process called Document Retrieval, which makes all the organization knowledge available via chat for anyone, getting detailed insights of systems, implementations, architectures, and designs, without even the need to ask the original authors, which may no longer be in the company at that time.

The Impact of AI on Software Development Teams

All in all, the speed of these releases indicates that in 2024, it will be, in line with Ray Kurzweil's vision, a ‘Singularity’ point, and the way companies and engineers are used to work will completely change. This will be extended to software engineering teams and the rest of the organizations since, nowadays, every industry is tech-savvy, and almost every company is becoming a software company. Think of automakers, electronics device manufacturers, food companies, retailers, and distributors; every business relies on technological support. A car, for example, requires software to be designed to be manufactured, and it contains up to 150 ECUs (electronic control modules) and, hence, is entirely based on software. Software platforms manage, control, and monitor Any modern manufacturing facility in real time. Retail usually requires operating around 50 to 100+ software modules, platforms, components, and interchanging information via complex integrations.

All this context indicates that the new Software Development companies will need to become more sophisticated than ever before. PM, Designers, Engineers, Testers, DevOps, and all the related roles will require a new mindset more oriented to the “What” than the “How”. AI came to help us with the “How” in ways we are still discovering and some that we cannot imagine. For example, tools like Copilot allow someone who could be more expert in a particular language to create great code and even learn without losing productivity, but most importantly, quality.

The Role of AI in Enhancing Software Security and Scalability

Since everything is software-based, scalability, availability, and security become a must; the supply chain, for example, is the engine of the modern world, and any downtime may produce a shortage in goods and services, which can stop operations of many companies or have a significant impact on consumers basic supplies. Software systems must be secure and provide robust protection against possible private data compromise, for example, and this is another area in which an AI tool can provide insights into potential vulnerabilities, analyzing an entire project. Tools like SonarCloud address this, but now LLMs can also perform a complete and detailed analysis of code quality and infrastructure security, providing detailed reports of vulnerabilities in the different cloud services.

The Changing Landscape of Software Development Skills and Education

With all these AI-based ecosystems, different areas of the organization will work together and iterate faster than ever to create and implement the software platform solutions required for their operation. Marketing, Sales, Finance, and Engineering are all interacting and in sync with Operations, getting solutions deployed in days or weeks instead of months, with the expected usability, documentation, quality, and productivity that modern business dynamics require. Software vendors will need more than ever to become active agents, interacting with all these stakeholders almost in real-time and providing solutions at a pace and with a quality that, in the recent past, we could only dream of. Software development Team members will be required to develop new skills, change their mindset, get more proficient with these tools, and work with the latest and most powerful programming language: English. Good prompting will be more required than being an expert in a particular programming language. Recruiting technical assessments will also need to change to evaluate these new capabilities. Now, candidates won’t be stressed by the typical live coding challenges, which typically do not provide real insight into a person's abilities, and instead, a candidate can be asked to solve a problem or even create a small project, something that can be done in minutes with the help of AI tools, and without googling for namespaces, or libraries, or any particular programming language syntax or keyword. This will make senior engineers more business-oriented, and experience in real-world projects and business will be more valuable than expertise on a particular tech stack. Universities and educational institutions will also need to incorporate these subjects into their content and become more focused and oriented to practical training rather than a theoretical approach.

Embracing the AI Revolution in Software Development

We are excited to see what is coming in the next few months and to adapt to this new context in the same period, a process that, in AccelOne, we started in 2022 and accelerated in 2023. We witness a new “Singularity” point in human history, even more significant than the Internet or personal computers. The main difference is the exponential nature of this process and its benefits for humanity. In the next couple of years, we will see new technologies and roles in our industry and how everything gets in place together to provide a better world for everyone, one that we need to be prepared for.