Get inspired - Use cases for your AI initiative

There’s so much buzz around AI but still very little understanding. It’s easy to get overwhelmed when thinking of your first project. You might be wondering: Where can I implement AI? Which use case should I start with? That’s understandable. One of the easiest ways for demystifying AI is to see examples of how you can use it to amplify your day-to-day work. This blog post provides you with some resources and things to consider before starting your first AI initiative. 

Get inspired

Try to think of a business problem that you’re facing and look at some of the existing use cases to get a practical understanding of what can be achieved; Dataiku’s Customer Stories or recent submissions for Dataiku Frontrunner Awards have plenty of examples of impactful solutions.

Have a quick reality check 

Creating predictive models requires a lot of relevant, clean data. Think of your data sources. What do they contain? Could they be helpful for the task? Is there data missing and would need to be collected specifically for the new project? How much data is there? While it’s difficult to say how much data you exactly need, more data usually improves your model’s predictive power.

Consider the time to value ratio

If you’re beginning to work with Dataiku and planning to develop the project yourself, focus on creating one quick, easy prototype to get a gist and iterating on it later. Quick wins are crucial to building momentum. I’m suggesting three use cases that work for almost everyone in another blog post.

If you’re interested in advanced use cases, take a look at Dataiku’s Industry Solutions. They contain customisable templates that help reduce time to deployment to a minimum. You can mix and match different projects and quickly adjust them to use your data.

The 3 ideal use cases for your first data science project

There are three use cases that prove especially popular with our customers. What sets them apart? First of all, they are relevant across industries, and their results are easy to understand. Secondly, you can use data already available in your organisation. There’s no need for additional efforts of data gathering. Finally, you can prepare them in no time with Dataiku’s AutoML. Let’s take a look.

• Churn prediction - Predicting which customers are likely to stop buying the company’s products or services during a defined time frame. Reducing churn is a straightforward way to protect revenue, especially for businesses with high customer acquisition costs. Later, you can combine churn prediction with uplift modelling to identify those most likely to be persuaded by a marketing initiative.

• Customer segmentation - divides the company’s customers into groups based on specific traits and factors they share. You can use segmentation for personalising targeted marketing campaigns aimed at customers with high churn risk. Personalisation also helps boost customer loyalty and conversions.

• Real-time fraud detection, e.g. on credit card transactions - a predictive model automatically flags transactions as potentially fraudulent, which can be either blocked or sent for further investigation. Machine learning approaches yield fewer false positives than traditional rules-based systems and better adapt to new fraud patterns. This leads to less regulatory risk, reduced human workload and better customer satisfaction.

Power up your BI with AI

If you’re a BI Analyst or a Citizen Data Scientist, you might have noticed a recent trend of augmenting BI with AI. It’s the next step in the evolution of a business intelligence solution.

Over the last decade, BI has become a staple in every organisation. It helped companies make sense of the massive amounts of data they collect by answering, “what has happened?”. But neat visualisations and dashboards looking into the past may not always be sufficient. So what’s next?

As BI maturity increases, the types of analytics evolve from descriptive to predictive. Predictive capabilities help us answer “what will happen next?”. It’s this capability that makes data science so valuable for business decisions.

BI solutions are not a one-time project; quite the opposite, they should continually evolve. A recent study led by Deloitte evaluated 152 cognitive projects and different strategies of implementing these initiatives into organisations. The results show that companies do better by taking an incremental rather than a transformative approach to developing and implementing AI, and by focusing on augmenting existing solutions rather than using revolutionary approach of replacing human capabilities with ambitious projects.

This means that implementing insight generated by ML into your existing analytics can be your first step towards AI. Sharing the results you generated with DSS can surface the value of predictive insights to more people.

If you’d like to learn more, check out Dataiku’s Power Visualizations with Dataiku + Tableau article and learn about the 3 main ways in which you can integrate Dataiku and Tableau:

1. Automating Tableau’s data pipeline with DSS and storing results in your preferred database

2. Exporting the dataset to .hyper file

3. Exporting the dataset to Tableau Server or Tableau Online with Tableau Hyper Export plugin.

How to become a Citizen Data Scientist?

There’s a lot of buzz around data science and machine learning. It might seem as though everyone has already jumped on that bandwagon.

As a consultant, I interact with hundreds of data professionals across industries and organisations, and to be fair; many enterprises have only just started their machine learning efforts. You’re not too late. The perfect time to start your own AI endeavour is now, and it’s not nearly as scary as it sounds.

Coming soon