Will AI Replace Jobs in Software Development Teams?
Manuel Schreurs
Helping businesses to scale up | Operational Excellence | Digital Change & Program Manager | Finance, Retail & Real Estate
The capabilities of artificial intelligence are growing. Many people wonder whether AI will eventually take over our jobs. My career focuses on helping (IT) organizations to optimize their operations, so I am curious about this topic. I have been exploring this question and testing the capabilities of AI, reading other articles, and talking with many people about it. And yes, I do like to provoke a bit.
While I love to think about how this will impact the entire world, let us start small: a core software development team of about 8-12 people.
From being skeptical to becoming a true believer
When I heard about ChatGPT, I did not expect much. I expected something like those horrible chatbots. I recently asked the Ikea chatbox: “Hi, I want to buy some shelves to put my books on. Do you have any?” It answered: “Hm, I couldn't find anything for that. Shall I connect you with one of my co-workers?” I was about to shout out: “It’s Billy, B-I-L-L-Y! It’s the best-selling product at Ikea and the most-sold bookshelves worldwide.” That was kind of my expectation of ChatGPT.
I have a solid technical background. I am well aware of the basic mathematical principles of AI. So I knew that it could be better than what Buddy did there. But when I tried ChatGPT, I was shocked. The results are so good, and the interface is easy to use. It was way beyond my expectations. And ever since, I have asked ChatGPT to do many things: to assist me in my research, my planning, and, yes, also in writing.?
I was curious how far I could take it, so I did a test. I asked ChatGPT to develop a simple calculator in React. I got all the code ready to deploy; pretty neat. But I thought that’s maybe just a standard answer ChatGPT took from an existing blog. So I asked to add specific functions like sinus or square root. It just did it all in seconds. Then I thought, let’s broaden the horizon and asked to write user stories and test cases. Amazing what it all did. It only refused to make the look & feel stunning. It was surprisingly stubborn suddenly. It looked like there was some prime directive in its core programming “Thou shall not make an app stunning.”
This small experiment has convinced me that AI will replace many jobs. The only question is when.?
Humans are expensive and slow; good talent is rare.
How would you decide if you could choose between an AI or a human to do a job?
Cost:?Suppose an AI can deliver the same quality as a human but at a fraction of the price. Well, then, the decision is easily made. Humans are expensive, and the current tax system focuses on making human labor even more costly.
Speed:?In the past twenty years, agility has been a hot topic in the IT industry. It is all about time-to-market. I experienced this firsthand when I had my own e-commerce business in China. I sold services and products online, like Groupon’s model. The competition was so fierce that I had to react extremely fast. So one day, a competitor started selling cinema tickets, and we knew if we waited too long, we would miss the boat. So within two days, we sold cinema tickets, too. I am proud of how quickly we could sign deals with cinemas, launch a marketing campaign and adjust our IT systems. It was not a small change. Users had to reserve seats during checkout, we had to send out SMS tickets, and we had to create a back-office system for cinemas to validate the tickets. This short time-to-market with a human team is possible but hardly feasible for any larger organizations I’ve coached. It would be a no-brainer if there were a way to adapt to market changes within seconds.?
Quality:?The cost of brand damage is enormous. Most of the bugs are caused by quite stupid development errors. Either created by a developer that’s too junior or by a senior developer that got bored of checking mundane stuff. AI doesn’t complain and can learn from all past mistakes.
Scalability:?It is no secret that good employees are difficult to come by. Having worked in scale-ups, I know from my own experience how this can limit the growth of the business. And many laws, especially in Europe, are focused on protecting employees, so scaling down isn’t easy either. AI doesn’t have any of these limitations.
But are we really that rational?
Besides looking at rational criteria, I also like looking at emotional ones. We, humans, don’t always make decisions entirely rationally. For example, as a manager, I may enjoy managing a team of humans because I want to say hi and drink a coffee or because it gives me a certain status. Or I may have a hunch that humans would do the job better. Or I’m afraid of new things, have ethical reasons that I don’t want to replace people’s jobs, or worry that we get too dependent on AI.
But the practical criteria will outweigh the emotional criteria. The practical benefits are too significant. Even if I agree with your reasons, companies will replace managers that do the same job faster, cheaper, and better.
Why do so many people disagree?
So I’m entirely convinced AI will replace software development teams. But many people tend to disagree. As a converted true believer, I call these arguments myths and debunk them.
Myth 1: Actual IT projects are much more complex than a calculator
I was a project manager at Capgemini for web content management projects. Whenever we wrote a bid, it had to pass risk management. In one of those sessions, the risk manager jokingly said: “Web Content Management, that’s just a database with a front-end, right? What can go wrong with that?” He knew it wasn’t that simple, especially if you talk about multi-country rollouts, but he was teasing me a bit. Yet it did bite a bit, as I thought he had a point. Most IT projects aren’t very complex. A big project can be broken down into smaller projects. Microservices are doing just that. Sure, there’s also very complex software, like when it uses a complicated mathematical model. I don't expect AI to develop advanced scheduling, video manipulation, or creating ChatGPT any time soon. But that’s such a small part of all the work being done. Most of the work is just a view on top of a data source.
Myth 2: AI isn’t that advanced yet; it can only do generic things
Sure, at the time of writing, AI can’t do what humans can. Sure, it can make a simple calculator, but there's still a vast leap to participate fully as a development team member. Or at least not yet. I believe that’s just a matter of time. I expect a variant of Moore’s law to apply here: AI becomes twice as intelligent every year. What will happen once AI starts writing software and writes the next AI? Sure, it may fail 1 million times and get it only right once, but it can do those million times very quickly. We are just beginning the AI revolution, and it’s already way ahead of my expectations.
Some people say it’s missing domain expertise. I do not understand why an AI cannot be trained in a specific domain. Such a specialized AI can read and remember many more books than a human can. And they can keep way more up-to-date with last-second developments. And if we only use that specialized AI in a specific domain, it can learn from all its users and evolve to become a domain specialist.
Myth 3:?AI is only a tool. It only does what you tell it to do; it will never manage humans.
I argue that computers already manage humans. Look at any warehouse: the order pickers see on their tablet which package to take from where to where. All their work is executing tasks a computer tells them to do.?
And who manages an AI? There are experiments of one ChatGPT chatting with other ChatGPTs. You give the big one a big problem, and it drills it down into subproblems. It then asks all the ChatGPTs precisely what to do and verifies their answers. Doesn’t ChatGPT become a manager here?
That AI is only a tool to be used by humans is something like saying: “A young professional is just a tool for a more experienced team member.” What’s the difference? Like in quantum theory: if you look at AI as a tool, you’ll see a tool; if you look at AI as a full-fledged team member, you’ll see a full-fledged team member.
Myth 4: We won’t let AI replace human jobs
This statement is wishful thinking. Whether you like it or not, we live in a capitalist age. If AI can replace humans, it will. Even if some companies resist, there will always be a competitor that will replace humans with AI and will become more efficient. It’s the same as the industrial revolution. At that time, people said that a machine could never replace a human and that authentic craftsmanship by humans would always be superior to machines. History repeats itself. There may be plenty of ethical reasons to disagree, but in the current set-up of our society, economic reasons prevail.
Myth 5: We’ll miss the human touch
You might say: “Software development is not just about writing code. It is also about understanding users' needs, and it’s missing empathy and creativity to design solutions that meet those needs.”
I love creativity, but I’m afraid that many software projects would be better off copying a successful recipe than trying to reinvent the wheel. I constantly emphasize the need to include end-users in requirements gathering. But honestly, I’ve also seen good POs draft better requirements in minutes than after a whole day of workshops. Like the famous quote that Henry Ford never said: “If I had asked people what they wanted, they would have said faster horses.” An AI can rely on almost infinite competitor cases cross-referenced with user statistics and reviews and even simulate them with other AIs. I expect the results will be better than asking the end-users.
Which roles will it impact?
AI will evolve. It’s a process. It’s not that tomorrow we replace all team members with AI. First, we will use AI as a tool. Then we will all become managers of AI agents doing part of our work: We will become more and more focused on orchestrating. The further AI develops, the more independent the AI can work. So step-by-step AI will eat away more parts of our human jobs.
What factors define how much AI can replace humans in software development?
As a good consultant, I developed a 2x2(x2) matrix to illustrate the level of impact per role.
The first dimension is orchestration vs. execution. Orchestration is about keeping an eye on the big picture, giving tasks, checking the work, and integrating it into the whole. AI is good at execution, but it needs somebody to give orders.?
The second dimension is the level of emotional intelligence required. Think about handling politics, change resistance, and reading between the lines.?
For now, I’ll also add a third dimension: complexity. But I believe it is just a matter of time before AI can handle more complex tasks than most humans.?
领英推荐
How will AI impact the job of Backend developers, Web developers, and Mobile developers??
AI will replace developers as we currently know them.?
Let’s be honest: most of the time, developers do mundane things. Like, create an API that reads or writes some data from a data store with an authentication layer. Or make a UI that shows data or allows you to enter it. Sure, they also write some complicated software, but I argue it’s about one percent of all work. And that one percent is challenging work; I have seen beautiful code with smart solutions and greatly respect the people that created it. But it's just such a small part of all the development work.
Their jobs will still exist in the first phase, but they will have to become experts and much more managerial. They will be telling AI what to do, and they will be checking their work and hence become 56% more efficient (Peng ea, 2023). They will also do the more difficult things that AI cannot do or teach AI how to do them. It is like how developing countries hire experienced engineers: first, get all the knowledge out of them, fire them later, and replace them with their own people. But in the long run, they are competing with AI, and eventually, AI will win. To stay relevant, they must move more to the right: become specialized managers managing pure AI teams.
How will AI impact the job of DevOps?
AI will replace DevOps.?
DevOps especially thrives on reproduction. It makes life so much easier if all environments are created similarly. Based on industry best practices. Developers will love it if all environments are set up precisely the same way. And doing this meticulously in a standardized way is right in the AI’s comfort zone. For example, an AI could create a new AWS instance, monitor system health, or adjust a pipeline to incorporate new tools - in seconds.
DevOps will have to become more like cloud architects, telling the AI what to do.
How will AI impact the job of Testers?
AI will replace Testers.?
Most test cases cover basic tasks like whether it does what the user story says. Another significant category is testing edge cases or unhappy paths. Sounds creative, but here I also see many recurring patterns. For example, you always try long inputs or strange characters. Again, I have relied a lot on good testers, and a good test manager is rare - but that is not the point here. The test cases I got for my calculator were definitely a good start.
Testers will focus more on risk analysis to create a relevant test plan and instruct AI on what to do. The role of DevOps and QA may merge, as so much is done by test tooling that needs to be set up and integrated into the pipeline.
How will AI impact the job of UX?
AI will replace UX.?
I love UX. But also here, I’ve seen so much repetitive work. I’ve often seen decision-makers that don’t like to copy so much as they think they’re unique. I would argue the opposite: UX typically improves if they copy existing designs, as re-used patterns are more recognizable to end-users. And an AI can also do UX research. Then, over chat, they can ask about typical use cases and create follow-up questions. You may even go one step further and create a new “User-AI” that observes an actual human for a few days. I wouldn’t be surprised that this “User-AI” can give better answers to the “UX-AI” than a human could.
Emotional criteria might slow this process down, e.g., managers being afraid to lose control or just liking to have workshops with real people. To stay relevant, they must step up their game and show their added value, using the creativity that makes beautiful software. I’ve met a bunch of very good UX’ers - they already do this, but I have also seen many designs that are not that great. The good ones will remain, and the lesser ones will lose their job to the competition.?
How will AI impact the job or Product Owners?
AI will take over part of the jobs of Product Owners but not replace them.
At first, I thought this might be more tricky, but then I asked ChatGPT to write user stories for a mobile banking app. I can tell you: the user stories were pretty similar to the ones I’ve seen in mobile banking projects, and I’ve seen a lot. But of course, it’s all about user interaction, so I asked which questions to ask to validate and improve them. Very similar to what I’ve heard in workshops. There will be a day when AI becomes proactive and asks questions instead of just responding to ours. As of that day, AI can fulfill a big part of the PO’s job.
The jobs will move more to the right or upwards in the chart. They become responsible for a more significant part of the chain. So they will have to take more ownership. Instead of creating an app that satisfies the stakeholders, moving to take ownership of the business objective of the app. It means they have to focus more on the product's positioning in the market. They can also move upwards and focus more on managing stakeholders. They become political players. An organization often consists of many silos with all different needs. They can create one harmonious, integrated vision out of that.
How will AI impact the job of Architects?
Architects will be one of the last to be replaced by AI.
I asked ChatGPT what architecture to use for such an app, but the answer was generic. For example, I did ask for specific libraries to use and patterns to implement, but it didn’t get very far.
This limitation is temporary. In the future, an AI can write a full Technical Architecture Document. But the real added value for this role is human interaction and orchestration. Software architects often work closely with stakeholders, including developers, project managers, and business analysts, to understand requirements, align technical decisions with business objectives, and communicate architectural concepts effectively. This collaborative and communicative aspect of the role relies on human interaction, empathy, and the ability to navigate complex relationships.
How will AI impact the role of Project Managers?
AI will replace the Team Managers but not the Program Managers.?
There is already a lot of debate about whether a PM is needed in agile teams. And to make things even more unclear, the role of PM is, in my experience, different in Europe, Asia, and the US. So let me separate it here into Team Managers and Program Managers.?
If the team is replaced by AI, the role of the Team Manager will be more administrative. And if it’s more administrative, it can be replaced by AI. As Gartner predicts, 69% of the manager’s job can be replaced by AI. (Gartner, 2020).?
The role of program managers is [in this context] the job of stakeholder management and making budget decisions remains. The more political side will be hard to replicate with AI. Stakeholders wouldn’t accept it - there’s no rapport between the AI and the stakeholder. And let’s also not underestimate the political games that will be played to be the last to be replaced - these people are masters at it. I expect their job will stay, even if only because we want to be able to hold a human responsible if things go wrong. As AIs can take over part of their job, they must manage more projects simultaneously. They also need to focus more and more on human interaction - becoming genuine change agents that make magic happen.
What’s next?
That does not sound like a very positive outlook, does it? So what’s next? Do we just let it happen, or can we take some action? Are we all doomed, or can the world be saved? As a good cliffhanger: stay tuned for my next blog!
I’m preparing a few at the moment, like:
As somebody who loves Agile's spirit, I will incorporate learnings from the comments below into my future blogs and reprioritize my backlog on your feedback.
Special thanks to Morne Wiggins for his valuable contributions and insights when I was writing this blog. I recommend everyone to look at his work at?https://agility-at-scale.com/.
About the author
Manuel is an accomplished digital transformation expert and leader who worked as a consultant, entrepreneur, and board member. He loves working with people and brings high energy and enthusiasm to every project he undertakes. With a proven track record of success in working with scale-ups and international companies and a strong focus on operational excellence, optimizing processes, breaking down silos, and driving organizational alignment, he is committed to helping companies achieve their business objectives and stay ahead of the competition.
References