November 24, 2025
Will AI make us smarter or dumber? The insights of Klöpping, Scherder and Online Dialogue
Reflection on Klöpping × Scherder by Simon Buil (Data Analyst at Online Dialogue)
From Jan. 29 to Feb. 2, it was time again for one of the highlights of the conference season for analysts: it was Superweek. I immersed myself in interesting presentations, passionate visitors, and enthusiastic hosts. The program featured a rich array of topics: from cookie tracking and privacy to KPIs and communications. Of course, as at many conferences lately, it was also a lot about AI. Because since AI has become increasingly accessible with tools like ChatGPT, more and more companies want to do something with it. Do you also want to start an AI project? In this blog, I share 5 lessons about AI from Superweek's speakers so you can use AI more successfully.
Julie Coquet and Ahmet Tarek of Media.Monks kicked off the conference and immediately came up with some important advice: make sure you have a clear question from the business before you start an AI project. Based on that question, you can then determine the potential impact. Determine what success looks like, and then address it as you would with a business case: “The expected uplift is x% per month, generating €y euros.” After that, you can start planning. Include in that not only the time needed to build the model, but also the time beforehand to understand the domain and infrastructure. Also remember that post-release maintenance costs time and money.
Ahmet Tarek on the Superweek stage.
To determine what success looks like, it is important to have the right KPIs to choose, said Michael Neveu, Senior Director of Machine Learning & AI Solutions at Media.Monks. With AI, large amounts of data can be processed faster and faster, and decisions are made based on that data. As a result, such tools can lead companies to both good and bad results at an ever-increasing rate. That's why it's more important than ever to really understand the problem you're trying to tackle. Think about use cases, what those should all include, and understand the models you are using. Based on that, set the right KPIs so that it is clear what you are working toward and can measure your success appropriately.
... Just a quick note: we send out a newsletter every three weeks that includes the latest blogs, team updates and, of course, news about the offerings in our academy. Click here to subscribe.
Ask-Y CEO Katrin Ribant showed in her presentation the importance of verifying the methodology and outcomes of AI models. She showed three models that she used Assistants in GPT-4 had created to predict which factors affect the conversion probability of visitors. To do this, she had imported GA data, then asked ChatGPT to create a predictive model. She repeated this three times, using the same data and the same prompt. The result was three models, with accuracies ranging from 61 to 68%. When she went to look at the methods of those models, however, she saw big differences. For example, the model with the highest accuracy turned out to be the user id for example, as a contributing factor, rather than as an identifier. So, even if the outcome looks plausible, always check how that outcome was achieved!
Katrin Ribant talks on Superweek about her prediction models created with AI.
Steen Rasmussen, Director of Data Innovation at IIH Nordic, came up with another recommendation: keep talking to your customers. AI is no substitute for contact with your target audience. Actually, this applies to everything you do with data: you can extract all kinds of insights from it, but you can only give meaning to it if you understand your target group well. And there is no better way to do that than by talking to your (potential) customers. In his presentation, Steen showed an example in which data showed on the one hand that Google search volume on pizza had decreased, but on the other hand pizza sales had increased. As it turned out, there is still a lot of search for pizza, but it's no longer through Google, but through delivery apps. And how do you find out? By talking to your users!
.... Do you want to know more about how best to keep talking to your customers? Or could you use some help with this? We can help you with user research!
Finally, there is the ethical side of AI. Privacy engineer Aurélie Pols stated in her presentation on privacy The following questions: who do you want to be? Someone who turns a blind eye when data is handled illegally or someone who only processes data that is on a compliant manner was obtained? Are you the whistleblower if you find out that more is being done with the data than is allowed under the contract with a particular party? There are privacy risks associated with any processing of data, especially if it is done with AI. Make sure you think about that, and you know what to escalate when and who to turn to.
Speakers at Superweek 2024 shared valuable insights on using AI. From the importance of clear demand from the business to choosing the right KPIs, and from controlling methods to continuing to communicate with customers, all of this advice matters for success. Focus not only on technological advances, but also on moral responsibility when dealing with data. A holistic approach is the key to sustainable success when using AI.
Want to know more about Superweek? Read the blog about Superweek 2023
Nederlands Nederlands
English