AI-first Tech Solutions for Scale-ups & SMEs

In one of the recent webinars, the team from Holycode shared a very honest look at what it actually means to become an AI-first software organization. The session was led by co-founders Laurent Decrue and Nenad Nikolic, who brought both business and technical perspectives from years of building and scaling products.

AI has been around for a long time, but something clearly changed when ChatGPT launched in 2022. That was the moment it became real for most companies. Suddenly, it was not just a concept or something in the background. It was something teams could actually use in their daily work.

The big challenge now is that AI is moving incredibly fast, while most organizations are not. Companies tend to change step by step, but AI is growing exponentially. This gap is where the risk sits. If you wait too long, you might not notice the impact at first, but eventually you fall behind in a way that is hard to catch up from.

Starting simple: where AI already works

Like most companies, the journey started with small and practical use cases. One of the first things implemented was AI meeting notes. These tools are easy to plug into existing workflows and immediately useful. They record conversations, create summaries, and even give feedback on meetings. For most teams, this quickly became standard.

Another example was AI outreach tools. These go beyond sending bulk emails. They look at a person’s online presence and generate more tailored messages. It is not perfect, but it works well enough to save time and increase output. These kinds of tools are a good starting point because they do not require big changes, but still show clear value.

From failed experiments to a real wake up call

Not every experiment was successful, and one good example of that was AI avatars. From a technical point of view, they worked surprisingly well, since you could generate videos and voice with very little input, but in practice they just felt off. There was something unnatural about them, which made them hard to use in real situations, and it showed that just because something is possible does not mean people will accept it. 

The real wake up call came shortly after with AI development tools, which had a much bigger impact. When non technical people started building working apps over a weekend, it completely changed how teams thought about software. If someone without coding experience can build something functional, then the barrier to building software is clearly dropping. This was a turning point internally as well, as it pushed teams to take AI seriously and start using it much more actively in their daily work.

Real impact on productivity and a shift in roles

Once AI was properly introduced into workflows, the results became very clear. Development became faster, with around 15 to 20 percent improvement, while testing saw even bigger gains. Tasks that used to take almost three hours could be completed in around 30 minutes using AI. Product discovery also improved significantly, as processes that included workshops, documentation, and prototyping became much quicker. Over time, this led to more than 50 percent speed improvement in some areas. A big reason for this is that AI performs especially well when it comes to repetitive and structured tasks, which is where most of the time savings come from.

At the beginning, AI was mainly supporting developers, helping them write code and move faster, but this relationship has started to shift. Now, AI is doing most of the work, while humans are guiding it. Instead of writing every line of code, developers focus more on setting direction, reviewing outputs, and making sure everything fits together. In one example, a full platform was built by just one person with some support, much faster than a traditional project and with solid results. This shift is important because it shows how the role of the developer is changing from being a builder to more of a supervisor.

What is coming next and where the bottleneck moves

The next step is moving towards fully autonomous systems. There are already experiments where multiple AI agents handle different parts of software development, from writing code to running tests and checking security, all while working continuously. The results so far are promising, but there is still a major challenge. It is not about whether AI can build something, but about whether we can control how it builds it. Questions around structure, code quality, and long term maintainability are still difficult to manage and require a lot of attention.

At the same time, the bottleneck in building products is starting to shift. In the past, execution was the hardest part, since building software required time, resources, and specialized skills. That is now changing. If AI can handle more of the execution, then the focus moves elsewhere. It becomes more about planning, making the right decisions, and getting products in front of users. Execution becomes easier, but knowing what to build and how to grow it becomes much more important.

AI: people, companies, and what’s next

One of the biggest challenges in adopting AI is not technical, it is human. Many developers were hesitant at the beginning because they did not fully trust the output and were used to having full control over what they built. Changing that mindset takes time. What helped in many cases was reframing the role. Instead of focusing on writing every line of code, the focus shifts towards guiding systems and checking results. In many ways, this is similar to how software development has evolved before, where higher level tools replaced lower level work and changed how developers operate.

From a company perspective, most are open to AI, mainly because it makes things faster and somewhat cheaper. The real challenge comes when old processes are kept in place. For example, traditional code reviews do not work well when AI is generating large amounts of code quickly, which creates friction between old ways of working and new capabilities. Companies that are willing to adapt their processes tend to see much better results. 

At the same time, the question of whether AI is just hype is no longer really up for debate. It is becoming a core part of how software is built, and companies that ignore it are taking a real risk. 

When it comes to jobs, there are still different views. Some expect job losses in the short term, while others believe the increased productivity will create more opportunities. Both can be true at the same time, but what is clear is that roles will change, with less focus on execution and more on planning, decision making, and oversight.

Final thoughts

Becoming an AI-first company is not about jumping on every new tool that comes out. It is about starting small, testing what works, and then building from there. It also requires a shift in how teams think and work, focusing more on guiding and supervising AI rather than doing every task manually. AI is not here to replace people; it is changing what people should spend their time on. That shift is already happening, and the companies that embrace it early are the ones that will benefit the most.

Access the full webinar replay in the Swiss Startup Association Education Library, free for members. Not a member yet? Join the community and get access to practical sessions that help you protect your business before something goes wrong. 

Don’t miss out on the latest news and events. Subscribe to our newsletter and stay up to date.