Hey Ruddi Rodriguez ! Thanks for your question. Enterprise ML platforms like AzureAuto ML, Sagemaker, Vertex AI, Datarobot, H2O etc. generally have a high security level compared to most of your own custom solutions. GDPR is still a roadblock for many companies in the the EU. I like to use these platforms for rapid prototyping because 99% of the more traditional businesses don’t have the resources to build and maintain their own custom ML stack. They need to focus on use cases and execution instead.

Hi Tobias , thank you for the answer , I have in mind that perhaps scaling and encoder , numerical and categorical data locally before to use a service in the cloud can be an option . Then the data that si exchange with the cloud does not contains data that can be tracked back.

Good approach to anonymize data first before you send it to the cloud!

Haha Ognen! No, it’s for people who are well versed with business intelligence (including the underlying data) and who want to improve their reports and dashboards with the help of some data science techniques. Nocode tools help them to use these toys and build fist prototypes without the need of looping in a data scientist or turning into a data scientist themselves.

Hi JaimeRV - great question! Prototyping is the way to go. In my experience, building a quick PoC in a small team (3-5 people) helps a lot as it demonstrates value even to non-technical management and get buy-in. It definitely helps if the management follows some top-level data or digitalization strategy as you can potentially link your prototypes/pocs to that which makes everything much easier.

Thanks for the answer Tobias Zwingmann!

Hi Ruddi Rodriguez, I chose Azure because I find it easier to use than AWS. Also, many companies build on the Microsoft stack for their BI (eg PowerBI) so I found Azure more natural for them. H2O is also a great platform. But even here, once you move to their enterprise MLOps platform you have to pay. Most non-tech companies are much better off taking an MLOps platform from the shelf and using it instead of trying to build with open source tools and reinvent the wheel imho.

Thanks for the answer, I see Power BI gives to Azure an advantage in many cases. So, if I understood well if I use AWS it lacks a tool integrated like power BI, sorry for the naive question I have zero experience using these tools I do everything locally mainly because of security restrictions. We are not even allowed to use power BI.

well “we” , I am the only one

AWS doesn’t have a native BI tool as far as I know. But it’s very easy to build a model with eg Sagemaker, deploy it as an HTTP endpoint and query this endpoint from within PowerBI / PowerQuery. That’s why I have not focused on the “native” integration between Azure and PowerBI in the book, but explained how to bring those tools together with a little scripting in Python or R.

Hi Tim Becker - glad to be here and thanks for your questions!
1- I had an aha moment once when I realized how dashboard users where using natural language queries very intuitively to get the metrics they wanted from a dashboard. It works really well. The technology has become mature enough and user have been trained to this kind of data retrieval over years because it works similar like a google search. I think key is when you really start with the basics and thus empower the masses. I have seen computer vision applications for turning unstructured data into structured data also deliver great results (eg document analysis)

2- An AI-powered analytics dashboard for me is when you use AI to improve your analytics/reporting experience throughout multiple stages (descriptive, diagnostic, predictive and prescriptive analytics)

thank you 🙂

Concerning question 1, do I understand correctly that the user would for example type “monthly sales” or “daily imports November” into a field and then get figures with corresponding data?

exactly! Even more complex queries work pretty well meanwhile thanks to the advancements in NLP. eg “sales in the US over time” or “daily active users january vs february as bar chart”

you can combine that with a custom business glossary so the tool speaks the language of your analysts which really makes it a powerful technology

The first example indeed is very powerful. But I think not many companies use are not using such tools in their daily tasks. Many companies are tight to Excel or even worst to delivering reports. Do you have any thoughts on why too many companies, including big banks, are still far behind in terms of AI? And why are they still struggling to change from power BI to tableau to excel and back without a good strategy to develop or adopt solutions as you mentioned?

Technology adoption always happens in phases.
Why did people still use typewriters when the could use computers?
Why did people still send letters when they could send an email?
Why do people still use email if they could use slack?
Sooner or later, new technologies will replace old, but it’s hard to predict which and when. We humans are crazy creatures 😃

But some general principles for enterprise AI adoption:

• must be triggered from top to bottom. No leadership buy-in, means no AI strategy
  
• you have to come up with a use case roadmap and tie that into a vision
  
• you need to sell that vision to the employees
  
• you have to acknowledge that culture eats strategy for breakfast

👍

Hi Allan, they won’t probably find it as valuable as the other way around. But if you’re looking for use case inspiration to apply your data science skills to BI (and especially PowerBI) it might still be useful for you.

Thanks Tobias, appreciate your taking the time to answer questions here!