In the latest episode of Growth Feed Podcast, our founder Sander van der Kraan talked to Lieke van der Plas about exactly that topic. Conversed.ai has been working with AI since way before the Chat GPT hype erupted in 2022 with AI and has built over 100 AI solutions for Dutch and International companies in the past 8 years. Sander talks about the most important lessons from practice, the pitfalls we keep seeing, and concrete tips for organizations that are serious about working with AI.
LLMs are not smarter than people
One of the key points from the talk: a lot of confusion around AI arises because people think AI is already smarter than humans. Sander explains that, in reality, a language model simply predicts the most likely next word. That may sound disappointing, but that down-to-earth explanation is the key to successful implementations. Those who understand what AI can and cannot do can determine much more specifically where the technology really adds value.
95% of AI projects fail
It sounds alarming, but it is anno 2026 reality: the vast majority of AI projects never make it to the finish line. Not because the technology doesn't work, but because of incorrect expectations and poor project management. Sander explains how companies often start from the technology rather than from a concrete problem. They underestimate the role of data and knowledge. And they forget that AI solutions require continuous maintenance and optimization.
AI is a tool, not a goal
Conversed.ai began 8 years ago with a clear conviction: AI should help businesses move forward in two ways, by saving money or making money. Anything beyond that is noise. Sander explains why that down-to-earth view is more valuable than ever in today's market.
Where do you begin though?
In the podcast, Sander shares concrete starting points for organizations looking to get started with AI:
- Start with repetitive tasks that your team loses energy from. That's where there are immediate gains, and that's where the return is quickly demonstrable. Then you can move to full agentic AI and tackle more complex problems.
- Start from a business problem, not a technology. What process takes too much time? Where do customers get stuck? That's the low-hanging fruit.
- Make sure your internal knowledge and data are in order. An AI agent is only as good as the sources it draws from. Conversed.ai offers a solution for this that automatically keeps knowledge articles up-to-date, so your AI is always working with the right information.
- Measure results in hard numbers. How much time was saved? How many questions have been solved? Without that yardstick, AI remains an experiment.
Managing expectations is leadership
Another key point from the conversation: AI projects often fail not because of technology, but because of leadership. Sander explains that organizations that do not create the right expectations within the company, among board, managers and employees, inevitably get stuck. Successful AI implementations require clear communication, realistic goals and the realization that AI is a journey, not a button you flip.
Listen to the full episode
The conversation with Lieke van der Plas is an open and honest look back at what Conversed.ai has learned in 8 years of building AI for businesses. For anyone wondering how AI can really add value to their own organization beyond the hype, this episode is worth watching.
Listen to the full episode on Spotify
Or watch the full episode on YouTube 👇
