Dieser Blogpost ist auch auf Deutsch verfügbar
When a handful of colleagues designed the first trainings on agentic software development a little over a year ago, it was completely unclear whether it would become more than an experiment. Since then, our trainers have delivered 47 trainings—on average, almost one every week—spread across 25 different companies. And demand is still high. Another 14 trainings are already firmly scheduled for the coming months, plus longer-term framework agreements with several customers.
What makes this year so special, though, isn’t the sheer number of trainings. It’s the pace at which models, tools, and practices evolve. Our team of trainers quickly learned that this training is never “done.” It requires even more continuous iteration than our other trainings. By now, they’ve found a good way to handle that.
In our first training, Claude Code shipped five updates within three days. These days, that doesn’t shock us anymore. We’ve gotten used to the new normal.
Ole WendlandSenior Consultant and Trainer at INNOQ
This constant evolution also challenges the companies we work with. Many of them therefore choose to rely on professional training to build up knowledge within their teams. That’s exactly where our training comes in.
Where our customers stand today
Many of our customers are in a similar situation: GitHub Copilot, Claude, and similar tools have long since been rolled out and are available to their teams. What’s missing is structured enablement, so everyone can get up to speed with these tools as quickly as possible. Knowledge levels vary widely as a result. Some individuals have already dug in deeply and are far ahead of the rest of their team.
For companies, the training is primarily about professionalization. Teams should reach a shared baseline instead of each person piecing the know-how together on their own. They learn it instead from people who work with these tools in real projects every day.
Compared to a year ago, our trainers see that participants today more often bring solid prior knowledge with them. They also notice that more and more companies aren’t just sending individual representatives anymore, but training entire teams. One customer even trained their whole engineering organization, voluntarily and organized internally, so everyone is ready for a day-to-day reality where agentic engineering is simply the norm.
Our trainer Fabian Walther describes what this does to participants like this:
Beforehand, many had only heard: here’s a tool, now you’ll be more productive. After the training, they feel like they can actually deliver on that promise. The structured approach gives them confidence.
Fabian WaltherSenior Consultant and Trainer at INNOQ
More than just tooling
From the start, our training was never meant to be a pure tool course. Instead, we wanted to teach the methods and practices behind agentic software development. That’s why we look at generative AI across the entire development process, not just when writing code. Which tool happens to be in use is secondary. That pays off twice over: new models appear monthly, and tools get weekly updates, yet the principles behind them remain valid even as the interface changes.
That’s also why requirements work takes up a lot of space in the training. Participants learn how to develop requirements, phrase them precisely, and hand them off in a way that agents can work with. The topic shows up in many places, from product work to quality requirements in architecture to practices like spec-driven development in implementation. Underneath are questions that sound trivial but are critical. Can I evaluate an agent’s output? And how does all of this fit into my day-to-day development work? Agentic work only delivers its real value once the individual steps interlock. A precisely formulated requirement, for example, is both the foundation an agent builds on and the yardstick against which its result can later be checked.
This also shifts your own role, from writing code to directing and evaluating. For many software developers, that’s the key takeaway from the training, and often a genuine eye-opener. One participant put it well: “So I’m basically a manager now.” Agentic work really does call for skills that used to be associated more with leadership roles. You take responsibility, delegate, and sign off on results you didn’t produce yourself.
This new role, however, requires understanding what you’re actually dealing with. What even is an agent? What can it do reliably, and where does it hit its limits? If you can answer that, you also know which tasks you can delegate with a clear conscience, and how carefully you need to check the results. This conceptual understanding is also what outlasts the next generation of tools. That’s why every training starts with the fundamentals, and of all sections, this one regularly gets the best feedback:
You might think the fundamentals are unnecessary. After all, everyone knows what an agent is. In reality, it’s the most valuable part. The feedback we hear again and again: thought this would be boring, but it was great to finally put all the buzzwords into a clean mental model.
Roman StranghönerSenior Consultant and Trainer at INNOQ
Snapshots from a year of training
After a year, certain moments stand out for our trainers: things they encounter again and again.
The mix matters
Heterogeneous groups work much better than you’d think. And that’s despite the training being aimed primarily at developers and software architects. When product owners, ops people, and security stakeholders are in the room too, everyone still takes something away in the end, even those for whom software development isn’t the core job. In one training, two analysts who had previously used neither Git nor Docker built a fully automated CI/CD pipeline within twenty minutes. That observation motivated us to adapt the training for other audiences too. With Agentic Work, we now offer a variant for non-developers such as product owners, product leads, or experts from business departments.
When skepticism fades
Almost just as often, participants show up who are initially skeptical about the topic. Over the three days, they realize they can suddenly try out their ideas within minutes, and that working with agents is actually fun. They often give the most positive feedback afterward, not least because our trainers are always clear about the limits as well as the possibilities.
When the exercise stopped failing
And sometimes progress overtakes the training right in the middle of it. In the early trainings, one exercise had participants let an agent build an entire application over the lunch break, completely and without intermediate checks. The exercise failed reliably, and at the time, that was exactly its didactic purpose. Then a new model was released, and in that very training, the agent completed the whole application over lunch without errors. That surprised not only the participants, but our trainers as well. No other moment shows more clearly how each new model unlocks possibilities that weren’t there before, and why we keep evolving the training.
Agentic software development is changing how people work. With our training, we lay the groundwork so our participants can keep pace with this change.
I’ve never experienced a training that changes participants' way of working as much as this one. For a three-day training, day 4 is always what matters to me: what can people actually take into their day-to-day work? And there, Agentic Software Engineering is by far the biggest change I’ve ever seen.
Michael SeelSenior Consultant and Trainer at INNOQ
Where the real work begins
As intense as the three days are, the real work starts afterward, when teams bring what they learned into their projects and processes. Companies can do a lot to make this phase succeed: provide the right infrastructure, explicitly allow for experimentation and mistakes, and give teams the time to stick with it.
Agentic Software Engineering was the trailblazer. By now, an entire family of formats has grown out of it. The foundation we laid there is really energizing a lot of people here.
Roman StranghönerSenior Consultant and Trainer at INNOQ
To support companies on this path, we’ve developed several offerings. The Agentic Engineering Accelerator Program picks up right after the training. We work alongside the team for a while, build up know-how and infrastructure together, and bring extra capacity so the first steps quickly turn into productive routine. Deep-dive workshops and talks address specific topics such as security or token costs. The security focus has since grown into a standalone two-day training: Agentic Software Security.
And then there’s the foundation agents build on. Clean architecture and well-structured processes provide the framework in which AI agents can operate at their best. That’s why it’s usually worth looking at your codebase before working with agents in earnest, and preparing it through targeted modernization. This is exactly where we feel at home. Software architecture and modernization have been at the core of our work since we founded the company.
Want to get your team ready for agentic software development?
You can find all information about our Agentic Software Engineering Training, the Agentic Engineering Accelerator Program, and our other offerings such as Agentic Software Security on our website. Feel free to reach out. Contact: Theo Pack, Principal Consultant at INNOQ.