Trainings

Category: AI

Insights from Netum’s AI Champion Training: Bringing Practical AI to Development Work

Posted on by Marjut Sadeharju

Splended had the pleasure of participating in and organizing an internal AI Champion training for Netum, where we guided a team of developers and specialists through the possibilities and practicalities of using AI in software development. The training was led by Marjut Sadeharju from Splended, with technical instruction by Aappo Pulkkinen from Forge Digital, and it turned into a development of AI-assisted coding and champion skills.

Exploring AI Through Code and Collaboration

Aappo describes the learning Spring with enthusiasm: throughout the Sprint, we dove into the use of various tools for AI-assisted software development and how they can support developers in everyday tasks. We explored the architecture and implementation of AI agents, experimented with Retrieval-Augmented Generation (RAG) models for more intelligent information retrieval, and examined the wide range of AI services and models freely available online. Participants had the chance to build, test, and iterate on real AI components, gaining hands-on experience in applying these technologies to real-world coding scenarios.

The participants reported their sentiment accordingly: it was exciting to see how accessible these tools have become. Integrating AI into software no longer requires deep expertise in machine learning. Instead, developers can tap into prebuilt models and APIs that offer surprisingly robust capabilities. And when those tools are explored together with colleagues, the learning experience becomes even richer. Discussions during the sessions often opened up new perspectives and revealed use cases that might have gone unnoticed in a solo setting.

Key Takeaways from the Sprint

One of the main takeaways was just how approachable AI has become. There are powerful, ready-made tools that can be integrated into development projects without needing to build models from scratch. This lowers the barrier to entry significantly and allows teams to experiment with AI even in smaller-scale applications.

Another important realization was the sheer volume and variety of AI models and services available online. From open-source models to commercial APIs, the ecosystem is already vast and growing rapidly. However, with this rapid growth comes a lack of established standards. We touched on this during the training, especially around technologies like MCP, which aim to bring more structure and standards into building AI systems.

A particularly eye-opening aspect was the cost structure of using AI via APIs. While tools like GitHub Copilot might seem “free” to the end user, there are also various different pricing models where costs can also scale based on usage, speed and model quality. It’s something that developers and organizations alike need to keep in mind when planning how to use AI in production environments.

Lastly, the theoretical side of AI turned out to be both fascinating and approachable. While the underlying mechanics of AI models can be complex, it became clear that a deep understanding of the math isn’t necessary for building effective solutions. Still, knowing the basics and understanding the logic behind the tools can give developers a much stronger foundation to build on.

A Shared Learning Experience

Perhaps the most valuable part of the training was the community that started to form around it. With a strong group of Netum participants bringing curiosity and experience to the table, we were able to build not just knowledge but also confidence. The Champion team is now ready to take on a systematic learning and development initiative, scaling learning within the organization.

Thank You to the Team

We want to extend our warmest thanks to the Netum team for their enthusiastic participation and openness to learning. It was a pleasure to facilitate a space where developers could grow their understanding of AI, and we’re excited to see how these insights will shape their future projects.

How to Help Product Owners Think Differently with AI

Posted on by Marjut Sadeharju

When we partnered with Riku Kuikka at Visma, the mission was clear: support Product Owners and adjacent roles in rethinking how AI could be used as part of real product work.

The goal was to create space to explore, question, and experiment. Because even in organisations that already use AI, there’s often a gap between knowing something is possible and actually building something new with it.

Coaching with Depth and Curiosity

To make that space meaningful, we brought in two of our trusted partners: Ville Lindfors (Codemate) and Aki-Ville Pöykiö (Adventure Club). Both bring not just technical fluency but deep experience in Product Development, systems thinking, and the art of asking good questions. Their role wasn’t to teach AI. It was to coach people into thinking differently.

Together, we designed and delivered sessions that challenged Product Owners to:

• Shift from “feature thinking” to experimentation thinking

• Understand the real boundary between automation and AI

• Explore how AI can support, not replace, judgment and creativity

• Prototype faster, and with a clearer sense of what’s valuable

Thinking in New Directions

One thing we appreciated from the Visma side was honesty. People shared openly where they were already strong and where the concepts felt either too basic or not yet grounded enough. The interactive sessions sparked meaningful conversations, also afterwards on their online channels and with our coaches, surfacing both excitement and skepticism.

That’s exactly where the best learning happens.

What resonated most:

• Practical examples that felt close to home

• Space for group reflection and discussion

• Clear coaching instead of static content

From Exploration to Execution

As the Learning Sprint progressed, a key shift occurred: thinking moved from experimenting to deploying.

“A great prototype doesn’t guarantee a production-ready solution. Getting to production requires versioning, risk management, transparency – and most of all, clarity about the AI’s role in the product.”

We explored what this means in practice:

• Ensuring privacy, legal compliance, and observability

• Logging and analyzing AI behavior in production

• Collecting user feedback and implementing A/B testing

Making AI-Compatible Products

An insight from Session 4 challenged how many participants viewed their systems:

“Our products don’t exist in isolation. They need to be AI-compatible.”

We’re no longer just building for humans. Modern systems should be accessible to AI agents as well. 

This means:

• Structuring APIs and documentation for machine consumption

• Considering AI personas in user modeling

• Preparing for interoperability with protocols like MCP (Model Context Protocol) and A2A (Agent2Agent)

In short: building for the future means designing systems that humans and AI can both work with.

Riku’s Vision Makes a Difference

What made this collaboration work was the leadership. Riku didn’t approach this as a box-ticking exercise. He saw it as capability-building: an investment in how product teams think, not just what they know. He brought the right people in, asked the right questions, and created room for the team to reflect.

Visma has set ambitious goals to record four AI experimentations per employee this year. With 600 employees, this means 2400 experimentations. Their top-level management is setting a concrete example on how they use AI.

“When an AI feature goes to production, it doesn’t just become ‘real’, it starts changing the entire ecosystem. Data flows shift, expectations evolve, and the way teams work begins to adapt around this new capability.”

It’s this kind of leadership that makes a difference. And it’s the kind of leadership we love to support.

Final Thought

If you’re wondering what real, grounded AI capability development for product roles looks like — it looks like this.

Learning Sprint on GenAI and LLMs: Valmet Developers Level Up

Posted on by Marjut Sadeharju

At Splended, we believe in the power of learning sprints to bring fast, impactful, and practical learning experiences to teams. Recently, we partnered with Nieve Consulting to conduct a specialized sprint on Generative AI (GenAI) and Large Language Models (LLMs) for one of our valued clients, Valmet. The coach for this sprint, Iván Moreno, is an AI specialist with experience in guiding tech teams through the latest innovations in AI. The Learning Sprint was conducted fully online,

The Challenge

Valmet’s team of 14 developers wanted to upskill in AI-related topics, focusing on gaining practical knowledge in GenAI and LLMs. With AI evolving fast, they needed more than just theory; they wanted to apply AI to real-world projects and feel confident using these technologies in their day-to-day work.

The Sprint

The learning sprint, led by Iván Moreno, was structured to give the team a comprehensive understanding of GenAI and LLMs while emphasizing hands-on learning. The sprint was tailored to meet Valmet’s needs, ensuring that the developers could:

  • Grasp the foundational concepts of GenAI
  • Experiment with real-world applications like Retrieval-Augmented Generation (RAG) in their projects.
  • Build confidence to integrate these AI tools into their workflow.

The Outcome

The feedback from Valmet’s developers was positive. Here are some of their reflections:

“I managed to get a general understanding of GenAI, which was exactly the target of the course!”

“I made a fun little magic: A Proof of Concept (PoC) with RAG!”

This feedback speaks to the effectiveness of learning by doing. The team didn’t just learn AI concepts—they applied them immediately.

Why Learning Sprints Work

This sprint exemplifies why we at Splended are strong advocates of the learning sprint model:

  • Focused content: By zeroing in on GenAI and LLMs, we ensured the team was able to understand relevant topics.
  • Practical application: Each developer had the chance to experiment and build during the sprint, making the learning experience stick.
  • Expert coaching: With Iván Moreno’s expertise, the team received tailored guidance and support, which boosted their confidence and competence in AI development.

Partnering with Nieve Consulting and seeing the success at Valmet reinforces our belief that hands-on, tailored learning sprints are the way forward in upskilling tech teams in cutting-edge topics like AI.