Artificial Intelligence in Analytics

We live in a fascinating time when Artificial Intelligence (AI) is transforming the way we do things. This includes data pipeline design and analytics. Today I’d like to talk about how AI powers automated big data analysis and reporting. I’m sure you are familiar with Business Intelligence (BI) if you are reading this post. Throughout my almost 15-year career in analytics, there was a persistent discussion about the impact of artificial intelligence on BI. It’s quite difficult to say what’s bigger: AI’s BI merge and its great potential, or all the buzz around it. This narrative reflects my personal views and beliefs regarding the evolving role of AI in analytics and Business Intelligence.

Companies aim to make better decisions based on the huge data volumes they collect every minute. BI as a discipline aims to analyse that data to generate insights that might have a certain monetary value. This in return grants a competitive advantage. However, BI effects on this are somewhat limited. That’s where AI comes into play bringing all the benefits of AI-powered enhanced process automation. So how does it work exactly?

AI-powered Business Intelligence. A hype or reality?

Is it just another ephemeral spark or is it going to change the way we do analytics?

AI/BI merge and its benefits

No doubt, BI is an important piece of any data platform design but it has intrinsic flaws that restrict the value it can provide to a business.

The analysis is an intrinsic BI task and BI’s primary focus was on data visualization for many years. The problem with it is that BI itself can’t predict data results and can’t create suggestions.

Predictive AI features

Take, for example, Sisense capabilities to predict trends [1]. It’s a very basic linear regression exercise incorporated into a robust BI tool. It’s a managed feature so analysts don’t need to worry about the regression model itself. BI tools does it all and its AI engine has the following models under the hood:

Prophet
Holt-Winters
Random Forest
Auto ARIMA

If we take AI as a process of teaching the machine such human capabilities as problem-solving and learning (the official definition) then it becomes obvious how potentially significant this impact on BI might be. AI-powered BI capabilities can remove those limitations. We can use AI not only to understand text but also to understand what users might want and to generate further suggestions.

Natural language

In many ways, AI helps business stakeholders find the information and insights they need. For example, in one of my previous stories I wrote that BI tools like Sisense (Simply Ask) [2] and Thoughtspot use AI-powered intuitive "Google-like search interface" [3] to get data insights from any modern data warehouse solution, i.e. Google Big Query, Redshift, Snowflake or Databricks.

Modern Data Engineering

In my opinion, enhanced business intelligence automation is the most significant way AI is changing the BI process.

Time is money and it is as simple as that. If AI helps to generate insights faster then it is a viable technological improvement worth further implementation into the BI process. Businesses that use Artificial Intelligence beat their competitors in terms of revenue generation and overall operational performance.

Another great example is Duet AI [4] integrated into Google’s Looker BI solution. With only a few sentences Looker users can create calculated fields, formulas, advanced reports and even more – create entire presentations in Google Slides.

Now getting insights from the data is as simple as a Google search.

Low-code integrations

Some tools go even further to enable not only chat-style data exploration but low-code or no-code at all ML model creation, management and deployment [5]. For instance, DOMO.AI offers an ML model management system with low-code interaction where we can connect an OpenAI adapter to import a model and configure and train it.

AI-powered explanations

Many BI tools will offer narrative and explanation features. How does it work? AI behind your report analyses all elements (fields or field combinations) that have caused the change in a specified point in a data collection (time, location, etc.) For example, it can identify that the device category is a possible explanation for a certain increase in revenue.

One of the most consistent examples of AI-BI pairing and collaboration is when AI can advise end users on what to do based on BI data. Imagine a marketing report with promotions and other campaigns. When the marketing team interacts with those insights AI can suggest how to optimize new campaigns for a particular audience or another promo event.

Segmentation

Another great example is how AI helps with segmentation in BI. Take, for instance, the clustering example when an AI-powered reporting solution generates suggestions on how to optimize retargeting for different user audiences. For example, with Tableau’s Einstein Discovery [6] enabled we will get recommendations for improving the predicted results. It swiftly runs through millions of rows of data to uncover significant correlations, forecast outcomes, and recommend strategies to enhance those expected results.

Conclusion

Business Intelligence, as a concept, refers to systems that automate Big Data processing and analysis. The full value of business intelligence is recognised when large amounts of data are broken down into granular insights but it has certain limitations as a process. It helps to understand a wider business picture in a nutshell. With AI getting insights from your company data should be as simple as querying Google. AI automation is a highly welcomed business feature as businesses focus their efforts on building their channels and cutting overhead costs related to a better user experience. Companies that use artificial intelligence beat their competitors in terms of revenue generation and overall operational performance. Despite this, many firms are still lagging when it comes to incorporating AI into their analytics. It is a noticeable trend in the BI world where almost every solution tries to offer AI capabilities by bringing key model-discovered features to the semantic model for feature impact analysis [6]. If you are a company executive it is worth reading a few books on AI-powered analytics.