Blog

Ramona: Martin, it’s great to meet you! I’m still new to the team, so I’m particularly interested in talking to someone who has been part of HiQ Germany for quite some time. How long have you been with the company?

Martin: Hi Ramona! More than nine years now. I joined as a Junior Software Engineer. Since then, I’ve seen a lot of exciting developments: new locations, a wide range of client projects and the merger with the Swedish HiQ Group. And today, I’m seeing how AI is shaking up the IT industry – and, really, the economy as a whole.

Ramona: Your background fits the topic pretty well. You studied Computational Linguistics, right?

Martin: Exactly. The intersection between people and technology has always fascinated me. I was never interested in talking only about the technical implementation; I also wanted to understand what products are actually supposed to do for people.

Looking back, my studies almost seem like preparation for today’s world of AI and LLMs. But, to be honest, that wasn’t really the case at the time. At least not in the way we understand AI today. Of course, there was already research into areas such as speech recognition, parsing and statistical methods. The first voice assistants, such as Siri and Google Assistant, were entering the market, but Large Language Models (LLMs) as we know them today were still easily another ten years away.

Ramona: Today, things are very different – it feels like everyone is working with AI now. So here’s the big question I keep hearing: Will AI replace software developers?

Martin: I think the panic around AI is overblown. I really do. What we’re seeing right now is a change in the nature of certain tasks, not the elimination of an entire profession. New roles are emerging, and new knowledge and skills are becoming necessary. That’s actually a classic pattern throughout the history of technology.

Ramona: What do you mean by that specifically? What is changing?

Martin: One thing AI is very good at is taking over dull, repetitive tasks – exactly the kind of work that intelligent people often find boring and that, as a result, can lead to mistakes. If AI can take some of that workload off our hands, that can be a good thing.

Human review, however, remains essential. What stays – and becomes even more important – is architecture, systems thinking and quality. Someone still needs to decide how a system can remain stable, secure and maintainable over the long term. Human review is central to that. AI cannot do this on its own.

From the perspective of an IT service provider such as HiQ, which works with a wide variety of clients, the focus is shifting. Purely mechanical development work will become increasingly automated, allowing us to place even greater emphasis on our high standards for system architecture, stability and quality.

The interesting questions are: How do we build robust, secure and long-lasting systems? How do we integrate AI in a meaningful way? And how do we make sure the result does not merely “work”, but can also be operated responsibly?

Ramona: Does that apply only to software development, or to other industries as well?

Martin: Of course, other industries can benefit from these new tools too. LLMs aren’t only useful for coding. They can also support areas such as leadership, process optimisation, efficiency improvements or onboarding, for example.

Ramona: So there’s nothing much to worry about?

Martin: It would be too simplistic to say, “It’s just a phase – in a few months, everything will be back to normal.” We can already see that many activities are changing.

New roles and skills are emerging, while other tasks are becoming less important or increasingly automated. This kind of progress cannot – and should not – be stopped. What matters is how we deal with it.

If all you do is mechanically work through tickets, things are likely to become more difficult for you in the future. But if you take responsibility, communicate with other people and understand their requirements, think about systems holistically and use AI deliberately as a tool, you’re more likely to become even more valuable with AI and agentic coding at your side.

These new tools are shifting the focus: they take work off our hands, but at the same time they demand a higher level of judgement and systems understanding.

Ramona: How is HiQ itself using AI in software development?

Martin: We’re looking very closely at AI-assisted and agentic software development, always with the goal of making development more efficient and more effective. The aim isn’t simply to bolt AI onto existing processes, but to integrate it meaningfully into modern software development.

At the same time, we regularly exchange knowledge internally at HiQ through workshops, experience-sharing sessions and other formats. We share what works, where the limitations are and which best practices are emerging. That helps us make sure AI isn’t merely something we experiment with, but something we integrate into our work in a responsible and meaningful way.

Ramona: Is it already working well?

Martin: The level of maturity depends heavily on the use case and the model being used. Our experience is that sometimes LLMs accelerate work enormously. At other times, they create new loops because results need to be reviewed and refined, or prompts need to be improved.

LLMs aren’t good at everything either, even if they can sometimes give that impression at first glance. That’s exactly why it’s important to stay realistic and not blindly follow every hype cycle.

LLMs are not a magic wand that simply makes complexity disappear. They can achieve a great deal, yes. But only if people guide them sensibly, critically review their outputs and establish clear parameters. Human review is essential.

Ramona: How do our clients view AI?

Martin: That’s exactly what makes our perspective as an IT service provider so interesting: we see what is actually happening across industries, not just what appears in press releases.

Some organisations are already very advanced when it comes to AI and agentic coding. They’re experimenting extensively, testing new models and making targeted investments. Other companies, particularly outside the IT sector, are still waiting to see how things develop.

Both approaches are legitimate. Our job is to understand where a company currently stands and, from there, work out the next sensible step.

Ramona: You’ve mentioned “human review” several times now. What exactly do you mean by that?

Martin: We assume that, in the long term, most – if not all – software projects will use agentic coding. The potential efficiency gains and opportunities are simply too significant to ignore. From my perspective, it’s not really a question of “if”, but rather of “how” and “how responsibly”.

Eliminating the human review process would be a mistake. LLMs are powerful tools and very good pattern-recognition systems that can generalise impressively. But they don’t truly “understand” what they’re doing, and they don’t take responsibility for the outcome.

That’s why they need careful and competent human oversight. Simply writing prompts is not enough. The people working with these tools need to be highly skilled.

And this doesn’t only apply to software development. LLMs can provide support in other fields as well, but they do not replace professional or legal responsibility. If, for example, an LLM is used to help calculate the structural integrity of a building, you still need a qualified architect who can review, correct and revise its results.

So, fundamentally, what is changing is the nature of the work.

Ramona: That sounds like a lot of responsibility. Suddenly, it doesn’t seem quite so simple anymore.

Martin: Absolutely. In many areas, you can see LLMs reaching their limits. One major risk is accepting AI-generated results too uncritically. Just because an LLM gives an answer confidently doesn’t mean that answer is correct or responsible.

That’s why I place so much emphasis on the need for human review. On closer inspection, the results often turn out not to be quite as good as they initially appear and require additional work or clear rules.

Future costs are another open question. And the issue becomes particularly serious wherever decisions can have a direct impact on human lives – in hospitals, aviation or critical infrastructure, for example.

Ramona: How can we make use of these new developments while still taking our responsibilities seriously?

Martin: We can’t focus only on the immediate benefits. AI is energy- and resource-intensive and has significant implications for infrastructure, supply chains and the environment.

That brings ethical considerations into the equation. AI is not a neutral tool. The way we use LLMs has consequences: for products, for companies, for people and for our planet.

I recommend that everyone engage with these questions. There is some excellent content by Alke Martens – for example on Instagram – who is Professor of Practical Computer Science at the University of Rostock and works extensively on ethical issues in computer science. Perspectives like hers can help us sharpen our own sense of direction.

Ramona: Listening to you, it sounds as though this isn’t really about technology at all. It’s much more about people.

Martin: Yes, and that’s no coincidence. Technology is exciting and important, but it isn’t an end in itself. Products are made for people. And ultimately, projects succeed because of people and teams.

That requires trust, openness and psychological safety. If someone makes a mistake – whether on their own or with the help of AI – they need to be able to talk about it without being afraid of losing face.

For me, leadership is primarily about relationships. Projects rarely fail because of technology. They fail because people don’t feel seen or respected. We build teams in a way that prevents that from happening.

When clients work with us, they aren’t just doing something good for their business. They’re also doing something good for the people involved in the project. That matters to me.

We use AI as a tool, not as a replacement for people. That difference is fundamental. We help clients build products for people.

Ramona: In what way?

Martin: We want AI to be used in a way that genuinely supports people and enables them to do good work. Not job losses through automation, but freedom from mundane tasks so people can focus on what matters.

Our message is this: When you hire us, you work with people who care about one another and who see AI as a tool, not as a replacement for people.

For me, that human touch isn’t a “nice-to-have”. It’s a clear success factor – both professionally and personally.

Ramona: Wow. We went from talking about your background to some very current topics pretty quickly. I think I need a moment to let all of that sink in. If someone reads this and thinks, “Okay, I want to tackle AI in my organisation, but I’m not really sure how to get started” – what would you tell them?

Martin: The most important step is not to start with the tool, but with the problem. Which processes take up a lot of time? Where are the bottlenecks? Where do errors occur? Where could teams benefit from additional support? And where can AI create genuine value?

The next step is to prioritise use cases, assess the technical and regulatory framework and start small – but do it properly. A controlled pilot project with clearly defined success criteria is often a good first step.

Business teams, IT, security and compliance should be involved from the outset and approach the topic together.

For me, this isn’t about immediately selling the biggest possible project. It’s about working together to identify the next sensible step for each individual company – professionally, with people in mind and with a clear view of the future of software development.

If you’d like to explore how AI and LLMs can be used effectively in your organisation, feel free to get in touch with us at any time: hello@hiq.de

Architecture: More “China Model” Than “GovCloud”

Contrary to what some may expect, the AWS ESC will not simply be an add-on to the existing AWS Public Cloud. From an architectural standpoint, it is more closely aligned with the AWS China model:

• Isolated control plane: The ESC has a completely separate infrastructure for management and authentication (IAM).
• No physical connection: There is no logical link between user accounts in the Public Cloud and those in the ESC.
• Service availability: Companies should be prepared for the fact that, at launch, the full portfolio of more than 200 AWS services will not yet be available. New features will likely be added to the ESC with a delay.

The advantage is clear: unlike US GovCloud (which requires US citizenship) or AWS China (which requires a local legal entity), the ESC is available to all European companies with the highest demands for data residency and operational autonomy.

The Strategic Decision: Public, Sovereign, or Hybrid?

Businesses are now facing a new strategic choice. The decision is no longer simply “cloud or on-premise,” but rather:

• Pure Sovereign: Full compliance, but without access to the rapid pace of new features in the Public Cloud.
• Pure Public: Maximum innovation, but login data and metadata may fall outside EU jurisdiction.
• Hybrid cloud model: Using both environments in parallel, depending on the sensitivity of the workloads.

The Technical Challenge of Duality

Organizations that choose the hybrid model will quickly run into administrative hurdles. Because Identity and Access Management (IAM) is strictly separated, this means:

• Identity management: Without a clear strategy, users would need to manage two separate logins. The solution is an external identity provider (IdP) such as Okta or Microsoft Entra ID.
• Double setup: AWS IAM Identity Center must be initialized and configured twice – once for the Public Cloud and once for the ESC.
• Infrastructure redundancy: Many resources need to be defined and managed twice.

The Right Update at the Right Time

Updates are only beneficial if they don’t cause disruption. In industrial practice, the opposite is often the cause: the shift starts at seven o’clock, but the machines are blocked due to a spontaneous update. Production comes to a standstill and frustration mounts. Security or performance update that are essential for your company thus become a nuisance.

With a professional F-OTA system, you can plan rollouts in advance. You define clear update windows, coordinated with shift schedules, production cycles, or maintenance times. Analysis of usage data helps you automatically select the ideal times. This keeps your systems secure and performing well – without slowing down your customers or their employees at crucial moments.

Convenience as Part of the Customer Experience

Today’s customers expect products to “just work” – without tangled cables, cumbersome service processes, or on-site visits for every minor adjustment. F-OTA makes this possible: devices receive updates in the background without the user having to take any action. 

For your customers, this means: 

• always the latest security standards,
• new features without new hardware,
• less downtime, less effort, fewer support cases.

F‑OTA thus evolves from a technical detail to a clearly noticeable comfort factor – and a building block of your customer experience.

Feedback that Really Makes an Impact

Many companies collect customer feedback, but few can incorporate it into their products quickly enough. Without F‑OTA, feedback often gets stuck on the roadmap for the next product generation.  

With a well-designed F‑OTA setup, you can translate complaints, requests, or suggestions for improvement directly into product improvements: a problem is identified, the firmware is adjusted, and the update is rolled out specifically to the affected devices. Customers see that their feedback is taken seriously—not in the next product generation, but in their current device. This builds trust and turns one-time buyers into loyal regular customers.

1. Rethink Business and Revenue Models in Terms of Customer Value

Price lists from the pre-cloud era are slowing down growth. Instead of one-time license or project prices, usage-based, results-oriented models are what count: pay-per-use, subscriptions, digital add-on packages.

Industrial companies that bundle services, data, and software generate predictable, recurring revenue – and retain customers for years, not just until the next machine purchase.

2. Build a scalable technology infrastructure

Monolithic legacy systems are the stumbling block for any digital transformation. In the next decade, your company will need:

• Cloud-First instead of server rooms
• Microservices instead of monster applications
• Clean, documented APIs instead of point-to-point interfaces

This will allow you to roll out new digital services in weeks instead of years – and your IT will grow with your business, not against it.

3. Make Software the Most Important Value Driver

Your customers no longer just buy hardware. They buy availability, transparency, and security. All of this is created by software.

Whether it’s predictive maintenance, digital twins, or self-service portals, the high-margin revenues lie in the digital functions surrounding your core product.

Those who treat software as the heart of value creation differentiate themselves where competitors are still discussing machine features.