With over a decade in AI and automation, Daniel has helped more than 80 Nordic companies turn emerging tech into real results.
As co-founder and former CTO of Zimply.AI, and with a background at Capgemini, he brings deep hands-on experience from LLMs, GenAI, agents, and RPA. His work has saved clients over 10,000 hours and counting.
Co-founder of Helicon Technologies, which he, together with his brother Erik, built into a team of 50 data scientists and engineers specializing in the energy sector.
With almost three decades of experience as a software engineer, Gustaf brings a broad and practical perspective across AI/ML, data, cybersecurity, and platform and software engineering.
With a background in finance and leadership, Erik has experience from scaling and operating international businesses.
Co-founder and CEO of Helicon Technologies, where he, together with his brother Gustaf, grew the company into a recognized player addressing complex challenges in the energy sector.
For too long, businesses have shaped their workflows around their tools.
The recent advances in AI make it feasible to build custom tools designed around the businesses' workflows instead.
AI agents reduce the need for rigid mappings and strict instructions. They can adapt to variability in ways traditional integration systems struggle to match.
This comes at the cost of predictability. If you account for that, you can introduce guardrails, constraints, and oversight to make the trade-off manageable.
Using AI can erode knowledge, machine-generated code may not be copyrightable, AI output is sometimes incorrect, etc.
Understanding the implications is essential for long-term success.
AI can make work feel faster and more productive. That feeling is not always accurate. The impact of AI should be measured through outcomes, not intuition.
Current AI can generate output that is hard to distinguish from intelligence. In reality, it is a system of mathematical transformations applied to vast matrices of numbers.
You will have a better time if you remember this.
Overwhelming hype increases the variance of reactions. Some move toward complete buy-in, others toward total rejection.
Both positions are misguided.
What is efficient for the business is not always better for the end-user. Optimizing purely for cost and/or speed can degrade end-user experience.
AI makes it trivial to generate code, content, and contributions at unprecedented scale.
Low-effort output creates noise, slows collaboration, and erodes trust.
Respect this and focus on work that deserves to exist.
Prompts and other inputs to public AI systems (e.g., ChatGPT, Gemini, Claude) are visible to (and possibly used by) the provider.
Never include confidential or sensitive information unless you have taken proper measures.
Not all decisions should be fully automated. Don't underestimate the value of keeping a human in the loop. Especially where errors carry real cost.
Most everyone have access to the same models, yet results vary a lot between users. Using AI is a skill that improves with knowledge and experience.
From managing context and framing constraints to decomposing problems effectively, evaluating outputs, and knowing when to refine versus move on.
Generative models can produce different outputs each time you make a request, even if the inputs haven't changed.
This variability is useful (diverse solutions, creativity, handling ambiguity, adaptation), but affects reliability.
Don't expect 100% reliability. Assume variability and design around it.
And don't forget that there's more to AI than generative models. Many approaches are fully deterministic and may be better suited when predictable outcomes are required.
When anything seems possible, it is tempting to reinvent it all.
But the highest ROI usually comes from doing "boring" things. Start by reducing manual steps, improving internal workflows, and assisting existing processes.