The Journey To AI: Get The Data Prep And Wrangling Right

Businesses that want to embrace artificial intelligence need to fight the first challenge, which is  getting the data preparation and wrangling right.

“Enterprises need to realise that even if you have the best AI model in the world but if none of the data that you are providing or giving is valid, how are you going to build to trust them all,” says Nelson Petracek, chief technology officer at TIBCO Software Inc. in an interview with Networks Asia.

“Enterprises need to fight the first challenge which is not a pretty challenge get the data prep and data wrangling right. And then from there once you have the confidence in the quality of your data then you can actually look to tackle what insights you can gain from the data.”

During the interview, Petracek further talks about issues surrounding AI, and how can enterprises take better advantage of the data they are generating.

We’ve seen the videos of robots working in Amazon warehouses. How far are we from the machine revolution? Do enterprises understand what AI means for now? Are they looking to achieve the right benefits and insights from their data?

I think it varies by verticals and by regions. I think some verticals have a better understanding of what AI and ML actually mean and what the applicability is.

There are enterprises that say “Yeah we need AI and we need ML” but when you ask them what that actually means to them, they can’t really tell you. And really a lot of it’s being fuelled by the fact that enterprises have been putting together big data solutions for quite some time and they’re not seeing value necessarily out of solutions.

And I think in many ways they are right but they just need to take a step back and first sort out what are the use cases that they’re trying to solve for it.

It was hard to put in AI and a ML solution in place if you don’t really know what it is you’re trying to look for or the use cases that you’re trying to solve. There’s a range of expectations when it comes to with this particular technology and what it can do for them. Because the technology is going to be around forever.

So how do we go from machine learning to AI?

It is usually finding the first problem that customers need to solve and that is one around data preparation and data wrangling.  So they have all this data at hand and be it as a data warehouse, big data storage and they need to sort out what the data is, how accurate the data is – data quality is crucial.

The first thing they need to solve is preparing the data or data prep. Then from there, depending on how well identified the areas that they want to investigate, you can then start to apply different supervised or unsupervised techniques to try and get better value out of that data.

Enterprises need to recognise the kind of tools they have in place that allow them to quickly and easily build these models, deploy these models and run these models at scale. They can then look at the results and determine where they can go from there.

It has to be an iterative process as well. Enterprises cannot just run it once and forget about it.

From our experience it is a journey. Enterprises need to fight the first challenge which is not a pretty challenge get the data prep and data wrangling right. And then from there once you have the confidence in the quality of your data then you can actually look to tackle what insights you can gain from the data. – Networks Asia