5 Brutal Truths About AI in Qlik Sense (Why Your Selections Suddenly Don’t Make Sense)

Read time: 2.5 minutes

The Only Way to Make AI Work in Qlik Sense:

1. AI Was Injected into Qlik Sense
• No connection between Output(s) and pipeline.
• Building the flow:
→ AI creates a structure, validates the data, then creates a model for your data that explores that model.

2. If you can’t select it, it’s not a part of your model.
• The associative model will not work with text at all.
• Turn your text into:
→ dimensions ( topics, categories)
→ measures ( scores, count)

3. Real-time AI is ruining Qlik Sense’s UX
• Dynamic reloads are creating slow and unstable applications.
• Separate layers:
→ AI - batch processing
→ Qlik - fast in-memory exploration

4. Without Validation, you will Lose Trust in Everything.
• Bad data will travel through the associative model at Lightning speed.
• To Prevent Bad Data from Entering the Model:
→ Add schema checks
→ Add standardized fields
→ Add QA Sampling

5. AI must improve the way you make a Decision, not just make your model look smart.
• Features do not give you insight unless it directly relates to a Decision.
• Make sure you know:
→ How are you going to make a Better Decision?
→ What is the Selection Criteria for making the Best Decision?

💡Key Takeaway: 

Qlik's AI does not work in a vacuum... it propagates. So if there is a problem in the pipeline, then it extends outwards throughout the BI application.