There is a lot of discussion about using AI in software development, but what does the change actually look like in an individual developer’s day-to-day work? Heta Björklund completed Gofore’s AI-Driven Developer learning path, designed for software developers and created by Gofore’s own experts. She shares her experiences of the learning path and explains how AI has become part of her work as a software developer.
Before AI tools could be used more widely in Heta’s customer project, their use remained fairly separate from the actual development work. She could ask AI general questions, but could not, for example, provide it with project code or other project materials.
That changed in the spring, when the tools could be used more concretely in her own work. Around the same time, Heta started Gofore’s AI-Driven Developer learning path, where much of the learning is based on hands-on practice.
“You only learn by doing. You can’t just read theory or high-level descriptions of how AI could work. You actually have to use it yourself.”
The training and Heta’s work also supported each other. She could try out things covered on the course in her work, while her experiences at work helped her better understand the content of the training.
Using AI helps you learn where it really adds value
Heta’s attitude towards AI has changed along the way. At first, she was also somewhat sceptical.
– Initially, I did wonder a little what we actually need this for and whether everything really has to be done with AI.
As she has used the tools, both their possibilities and their limitations have become more familiar. At the same time, Heta has found situations where AI makes her own work easier.
The most useful tasks are often those where the end result matters, but where the details do not need to become part of the developer’s own specialist expertise. When writing unit tests, for example, Heta can describe the test case, the conditions and what she wants to verify to the AI. Instead of spending time remembering the syntax of different languages, she can focus on the test case itself.
– I’ve liked using AI for things I’m not particularly good at and don’t actually need to be either. Things I know I’m good at are often easier to do myself.
The key is learning to recognise when AI genuinely adds value to the work and when it simply makes more sense to do the task yourself.
A good result often takes time and experimentation first
One of Heta’s most concrete experiences relates to the databases in her customer project. The project had four separate environments, each with two databases. The task was to go through a large amount of information and determine which old materialised views were still in use and how they were connected to each other and to other database tables.
The work could have been done manually, but Heta estimates that it would have taken days. Instead, she spent a couple of hours building, testing and refining a good prompt. Once she had found an approach that worked, she could repeat the same task in all environments. After that, it took the AI around ten minutes per environment to make an analysis and create a chart of the database structures and a list of views no longer in use.
The benefit of AI was easy to see in this case, but the working solution did not emerge from a single question. The task and the instructions first had to be tested and refined. Once a good approach had been found, the same analysis could be repeated with AI in all environments.
Learning to use AI therefore also means learning how to define a task in a way that produces a genuinely useful result.
AI suggests, the developer evaluates
The more AI does, the more important the developer’s ability to evaluate its work becomes. In Heta’s view, one of the key responsibilities of an expert is to understand why something is being done, who it is being done for and what really matters to the customer and the product.
AI also does not necessarily tell you when the work is finished. It can always suggest one more improvement or another thing to do next. It therefore remains the developer’s responsibility to assess the bigger picture and decide when the result is good enough.
At the same time, developers need to understand what AI is allowed to do. Heta highlights an experience where an AI tool independently tried to go looking for deleted files in a location it did not need to access. The experience made it clear why the instructions, permissions and boundaries given to agents need to be considered carefully.
Using AI does not, therefore, reduce the need to understand your own work. In many situations, the opposite is true: it requires developers to make even more conscious choices.
Skills also grow together
In Heta’s project team, AI skills are not developed only individually. The team has shared instructions for AI in a common repository, which they develop together. When someone finds a useful new skill or way of working, for example, others can start using it too.
One example is a prompt template the team created for assessing vulnerabilities. Based on the description of a new vulnerability, AI can use it to help determine whether the vulnerability affects their system and what should be done about it. In this way, an individual experiment can quickly become a way of working that benefits the whole team.
On the AI-Driven Developer learning path, peer support was also available through a shared Slack channel for participants, where they could ask questions, share their own insights and feedback, and discuss things with others. Heta found this particularly useful when she encountered technical problems during the exercises.
The learning path provides structure for learning something new and an opportunity to experiment. Your own work, in turn, shows which approaches work in practice. And when effective ways of working are shared within the team, everyone does not have to solve the same problems from scratch.
Next, Heta wants to explore more autonomous agents
After completing the learning path, Heta already knows what she would like to learn more about next. She is particularly interested in genuinely autonomous agentic workflows, where a task could be divided between several agents and progress further without constant human guidance.
However, not everything can yet be experimented with in her own customer project. AI cannot, for example, be given access to all the systems whose data an agentic workflow would need. The learning path has therefore also given Heta the opportunity to explore ways of using AI that are not yet possible in the day-to-day work of her own project.
In consulting work, it is not always easy to find separate time for learning something new. In Heta’s experience, however, the best situation is when learning and your own work support each other. The AI-Driven Developer learning path provided one structure for this.
“It was really great that the employer offers training like this. A lot of time has clearly gone into it, and the training was well delivered.”
Ultimately, Heta’s advice to another software developer looking to develop their AI skills is simple: use the tools yourself. Not everything needs to be done with AI, but you only discover your own ways of using the tools, and their limits, by experimenting with them. The goal does not need to be to use AI as much as possible. What matters more is learning to recognise when it is genuinely useful.
Are you interested in Gofore as an employer?
Grow with AI
AI-Driven Developer is an 11-week learning path designed for Gofore’s software developers, with a strong focus on developing AI skills through learning by doing. Around 70% of the learning is based on practical activities, with the remaining 30% focused on theory. The content has been created by Gofore’s own experts, and participants receive a certificate upon completion.
This article is part of Gofore’s Grow with AI theme, where we explore how artificial intelligence is changing expert work and share Gofore employees’ experiences of how their work is evolving and how they are learning new skills.
Read also the previous content from this theme:
Joonas Korpela’s blog What does expertise mean in the age of AI?
Article How can software developers keep their skills up to date in the age of AI?
