From Developer to Agent Coach: What Have We Learned from Agent Development Training?
06 | 2026 Jari Huilla, CTO & Partner at Kipinä
For about a year now, we have been organizing workshops and training sessions for our clients on artificial intelligence and agent-based software development. These sessions have provided us with a wealth of insights that we might not have even noticed in our day-to-day work with agents. Check out our top tips on how to integrate an agent-based development approach into your organization’s operations.
What AI skills are most important for organizations to adopt?
The two key pillars of agent-based development that generate the most "aha" moments in organizations are:
💡 The ability to imagine new things: Today’s tools are already so good that a developer’s imagination may be the biggest limiting factor. As with everything else, the best way to expand your imagination is to see a practical example of someone more skilled than yourself. However, your imagination may also be limited by fear if you haven’t mastered the basics.
🔐 Cybersecurity: When cybersecurity issues related to AI agent development are discussed clearly and without sugarcoating, the impact is often shocking. Did you know, for example, that an AI agent used carelessly can operate with the same access rights as you do on all the online services you’re logged into via your browser? Or that when a coding agent runs a command-line script, it’s important to understand where the script is actually being executed and what it might have access to?
How have our own thoughts changed over the course of the year?
Just a year ago, in the spring of 2025, the main message about AI in software development, put simply, was:
"Wake up! Language models have advanced so much and so quickly that you can actually use them to code."
Who would have thought how much things could change in a year?
By the summer of 2026, most software developers will already be aware of the technology’s capabilities. The discussion has shifted to how humans and AI agents work together, as well as how to resolve bottlenecks in decision-making and teamwork in a situation where the speed of software development is no longer the biggest limiting factor.
It has been a huge leap forward. At the same time, we have had to continually adapt the content and approach of our training programs to keep pace with a rapidly changing reality.
What are we paying special attention to now that we didn’t need to think about a year ago?
Just a year ago, most developers were very actively "managing" their code agents. The default assumption was that a person would approve at least tool calls and often file changes as well.
We are now paying much more attention to the paradox that agents are capable of carrying out significantly larger-scale actions on their own—for better or for worse.
If an agent can work independently for hours on end, who really wants to keep approving tool requests every few minutes? And who has the energy to genuinely evaluate each request carefully?
At the same time, agents are becoming increasingly skilled at using tools, combining them, and sometimes even circumventing the restrictions placed on them. For this reason, the most important factor is no longer constant human monitoring, but rather the ability to build environments where access to API keys, secrets, and other critical resources is structurally prevented.
Today, we have a much better understanding of the limits of human vigilance than we did just a year ago.
Agent-based development is, above all, about changing people
The biggest lesson we've learned this year is that agent-based development isn't primarily a matter of technology.
The tools are already capable enough. The real competitive advantage comes from how well an organization can envision new ways of working, utilize agents as part of teamwork, and build secure operating models around them.
At the same time, the role of the developer is changing; instead of writing code, an increasing portion of the work involves setting goals, directing agents, assessing quality, and managing the big picture.
If you want to understand what this change means for your organization specifically, check out Kipinä’s agent-based development training courses. We train development teams to utilize agents safely, effectively, and in a controlled manner—from the first AI assistants all the way to multi-agent workflows and building your own agent-based development environment.
Jari Huilla, CTO & partner Kipinä
Jari Huilla is Kipinä’s CTO, with an exceptionally long and diverse background in technology; he landed his first job at the Nokia Research Center at the age of 15. Over the years, he has worked as a developer, manager, and builder of growth companies. At Kipinä, Jari brings together deep technical expertise and business-oriented thinking. He is particularly interested in how to make AI solutions not only technically functional but also genuinely meaningful—and how to understand their limitations, rather than simply ignoring them.