The question I ask myself is no longer "What can we do with AI?", because if you ask that, you will always find something. For me, a use case almost always starts with a different question: "Where are the frictions in the process today?"
Today, I had the opportunity to show AquaFlanders what that can look like in practice: a solution that intelligently groups and summarises customer interactions from a call centre, so helpdesk employees can understand more quickly what has already happened and what the current status is.
That doesn't work by simply throwing a large amount of text into a single prompt. The interactions are first cleaned up and interpreted, then compared semantically and grouped by topic. Only then do we create a summary for each relevant group, with factors such as recency, relevance, context and the latest action helping determine what is ultimately shown.
So no AI for AI's sake, but making existing information more useful at the moment someone needs it, integrated into the process.
It was also great to exchange experiences with AquaFlanders, Pidpa, Farys, water-link and the other attendees. Different organisations, but remarkably many familiar challenges.
And yes: speaking in front of a room is still slightly less relaxing for me than building slides and AI models. 😅 But: survived, got the message across and had some very good conversations.
Thank you to AquaFlanders for the invitation and the organisation!