Another month has flown by. In this newsletter: happy news on Bram’s internship and a little bit of musing about process data preparation and AI.
Progress on the E and L of ETL
At Konekti, we’re all about speeding up the ETL process for process mining: Extracting, Transforming, and Loading process data into process mining tools. But truth be told, we focus mostly on the T of transformation. By streamlining the process of transforming process data into an event log, we’re able to save a ton of time on a process mining project.
But, as we all know, without the E of extraction, there’s no data to transform. That’s why we’re thrilled to announce that our intern Bram has successfully built an integration with an open-source extraction tool. This means our customers will be able to connect to hundreds of different systems – not just the two connectors we currently have. Unfortunately, Bram’s college doesn’t give grades for internships but trust us, he’s a straight-up 10/10 in our book. We’re beyond excited to have him continue working with us as a developer
Competition from the mess-AI-ah
If LinkedIn is a reflection of what’s going on in the process mining business, it seems like we’re all obsessed with the ChatGPT algorithm. Admittedly, we contributed to the stream of posts along the lines of ‘explain process mining in a simple manner’. But honestly, the real question is: can Konekti expect competition from ChatGPT and other AI innovations in the future? Will AI be able to take over the job of preparing data for process mining?
On the surface, ChatGPT seems to be doing remarkably well in turning English questions into working data transformation queries. Google ‘ChatGPT SQL’ and you’ll find several blog posts on experiments where it is shown that ChatGPT needs very little input to write code that is sometimes remarkably correct. Sometimes is a keyword here, as it’s only right about half the time. And we all know that half the time won’t cut it; trust in process data is hard to earn and easy to lose.
One of the hardest things about automating data preparation is that process data is often so ambiguous. Transactions and accounts mean different things in different contexts. And which dates are meaningful activities, and which are not relevant or not activities at all, but deadlines or target dates?
So, while we don’t think we can automate process data transformation, we do believe that there are ways to make the process of data transformation faster and easier:
- ? With a structured, no-code solution, process data transformation can be structured and simplified.
- ? If we’re willing to share our knowledge in a community, we can re-use existing work instead of being forced to start from scratch.
And with that, we went full circle and are back at what we started with: Konekti!
Are you curious to see what Konekti can do? Konekti’s product page is now live, explaining some of its most important features with screenshots. And for those that want to see it in action, there’s a demo button on the website as well, where you can schedule a demo with Maarten.