It is not about cheaper development – it is about higher ambition
We began making a serious shift in late 2024. Not because a plan was already in place, but because it became clear that something fundamental was changing in our craft – something we couldn't simply wait out. I have been in this industry long enough to have seen a number of "this changes everything" waves come and go without actually changing all that much. This is not one of those times. For the first time in a long while, it didn't feel like just another new technology to evaluate, but rather the start of a shift that is genuinely transforming how systems are developed. A year and a half later, I know a lot more. At the same time, I remain humbled by the rapid pace of development and uncertain about where it will all land. Yet, enough has become clear for us to articulate our perspective and explain why we are changing both our way of working and our models for collaboration.
Expectations are raised
Almost every discussion about AI centers on efficiency. How many hours can be saved? How much cheaper will development become? How many developers will be needed in the future? These are relevant questions, but I believe they risk obscuring the true shift taking place.
The major change isn't that building the same systems as before becomes cheaper, it is that expectations for digital products are rising. When everyone gains the ability to develop richer, faster, and more well-conceived solutions, that becomes the new standard. Your customers won’t compare your system to how things looked a few years ago, but rather to the best digital experience they’ve ever had – whether that experience comes from your industry or a completely different one.
That is why I believe companies do themselves a disservice if they view AI primarily as a tool for cost savings. The companies that stand to gain the most are likely those that use the technology to raise the bar – not to build the same solutions more cheaply, but to create solutions that would previously have been unfeasible to develop.
The developer's role is changing
One of the most common questions we receive is whether AI will write all the code. We believe the answer is no. Manual, human coding and oversight will retain their place – specifically in systems where the need for control is paramount. It is a matter of balancing cost and benefit, a decision that must be made deliberately, system by system – not as a matter of principle, nor by the vendor alone. There will continue to be systems where human coding is the right approach, and others where AI can handle a significant portion of the production.
However, we are absolutely convinced of one thing, the understanding of how code and solutions are actually built determines how much value you get from these tools. AI does not make experience less important; quite the opposite. Those who derive the most value from AI-driven development are the ones who already possess a deep understanding of systems development. They know what to ask for, how to ask for it, what to question, and when a result looks convincing yet is actually incorrect.
We therefore foresee a shift in the focus of our work. More time will be spent on solution architecture and advisory services, and less on the actual production of code – which is increasingly being handled by AI in systems where that is acceptable. Defining the right problem is becoming more important than writing every line of code oneself. Consequently, we are restructuring our organization so that more of our developers work closer to the client and take greater responsibility for the big picture. Our strength lies in our ability to operate across the entire spectrum – from architecture and requirements gathering to coding, quality assurance, and operations.
New collaboration opportunities
As the pace of development changes, the way we collaborate needs to change too. Hourly billing has long been a standard approach to purchasing system development, but as productivity rises, the number of hours worked becomes an increasingly poor metric for the value of the partnership. What we are truly agreeing upon is what needs to be done, in what order, and at what pace.
We want to place this agreement at the heart of our relationship; this is the rationale behind our new retainer model – a subscription-based arrangement with a fixed monthly fee, where the level is determined by your desired delivery pace. This model is paired with a working method that assigns you a dedicated solution architect to drive development together with you. Someone who gets to know your platform, your business, and your team, and who remains with you over the long term. We are able to offer this model because AI allows us to get more out of every hour worked. We believe this gain creates the most value for you in the form of a faster pace at a predictable cost. Furthermore, it saves both of us from spending time estimating and reconciling individual work items – time that can instead be devoted to actual development. The service levels range from a smaller commitment for those seeking a reliable partner and steady development, through a tier involving active product work, to a fully dedicated team for clients whose digital platforms are central to their core operations.
Fixed-price and time-and-material projects will still be an option, but we believe many will appreciate this new form of collaboration. What I observe among clients already using this approach is a shift in the nature of our conversations, the focus moves away from the cost of the next modification and toward which change will generate the most value next.
We continue to take responsibility
Two questions almost always come up when we discuss AI with clients. The first concerns responsibility. Who is actually responsible for code generated with the help of AI? For us, the answer has been clear from the start, if we deliver a system, we are responsible for ensuring it works. Our guarantees apply in exactly the same way, regardless of the tools used during development. Our quality control remains in place. It has undergone significant changes – not least because we use AI there too, for review and testing, but a human still reviews the code, and a human stands behind the final delivery. AI changes how we work, but not our responsibility.
The second question concerns data. We have chosen to work with market-leading models, currently primarily Anthropic’s Claude. We use subscription plans exclusively where neither our data nor our clients' data is used for model training or debugging. In addition, we have built a proprietary solution featuring a local language model and dedicated GPU power that inspects every request before it leaves our network. If a request is deemed to contain personal data or other sensitive information, it is anonymized before anything is transmitted further. This allows us to utilize AI even when handling production-grade data, without risking leaks. We believe this type of control will become increasingly important and will soon be a standard component of professional AI usage.
This is just the beginning
Ultimately, the decision regarding the extent of AI usage and which subsystems it should be applied to rests with you. Our role is to provide an honest assessment of the impact on pace and opportunities, the implications for control, and the associated costs.
These are undoubtedly exciting times. The entire industry is undergoing a transformation, though no one knows exactly what the outcome will be. The interesting question is how we choose to seize this opportunity. We believe that the most successful companies will use AI to raise their ambitions rather than simply cut costs.
We would be happy to assist you on that journey.
Would you like to explore what this could mean for you? We offer a workshop where we review your systems and opportunities together – to see what is relevant to you.
