5 Reasons Why AI Could Change The Way We Do Analytics

How Large Language Models have the potential to change the way we interact with data and how you can prepare now.

Read time: 7 minutes

Hey there,

Today's issue will be a special edition consisting of three sections:

  • A quick recap of 2022 with regards to AI,

  • my personal predictions for the next year, and

  • suggestions for concrete actions you can take now.

I hope this will give you a better understanding of the current and future AI landscape and identify what you can do to be better prepared for the changes and opportunities ahead.

Let's go!

Quick heads-up: The winners of the book giveaway were announced last week.Take a look at the comments in this post:

The Year of Generative AI

2022 will be an unprecedented year for AI. It could very well go down in history as "The Year of Generative AI."

Technologies such as DALL-E, Stable Diffusion, and most recently ChatGPT are just a few of the new breakthroughs that have amazed not only industry experts, but the masses as well. ChatGPT attracted 1 million users in just 5 days - breaking all other records for user acquisition for AI software.

And we're just starting to get into generative AI, as this article here shows:

I'm particularly excited about applications involving text inputs that are processed by Large Language Models (LLMs).

What will the next few years bring?

Are we entering a text-to-everything era as Armand Ruiz suggests?

My guess in terms of analytics and BI is that Large Language Models will forever change the way we think about dashboards and BI.

5 Reasons Why Large Language Models Could Change The Way We Do Analytics

Before I give my reasons for this prediction, I want to make one thing clear:

The use of NLP in analytics is anything but new.

For example, Microsoft already introduced the Q&A feature in 2019, which allows users to naturally interact with their data and dashboards by asking, for example, "What was last month's revenue in the U.S. compared to Europe?" This makes reporting much more user-friendly overall, and the feature has been steadily improving over the past few years. "Smart Narrative" - another NLP feature in Power BI - is also already helping users by automatically creating descriptive text for their charts and highlighting interesting patterns and trends.

But: I believe LLMs will take this experience to a whole new level.

Here's why:

1. LLMs got much better and more efficient

Already today chatGPT can answer complex questions about data you provide in the chat prompt. This makes answering questions much faster and easier than before. GPT-4 is already waiting in the wings and as I mentioned in this post here, it's far from the only LLM out there:

2. Fine-tuning will allow organizations to adapt LLMs to their custom domains

LLMs so far work very well for use cases involving publicly available data or common sense knowledge. They lack the ability to understand very domain- or industry-specific context or jargon. To solve this problem, LLMs can be customized or fine-tuned, as OpenAI demonstrates. With these capabilities, organizations will be able to create their own customized version of LLMs based on fine-tuning existing models to perform better on their own datasets and business glossary.

3. Most company data is still unstructured

To get an idea of the sheer volume of information stored in the form of text documents or emails, just think about your own company's data. Most companies have vast amounts of unstructured data that is difficult to process and analyze. That's where LLMs come in, as they can process this data at scale and extract meaningful insights from it. Instead of doing a sentiment analysis of user reviews, we can ask, "What are the most common issues my customers complain about?" and get an accurate answer.

4. Dashboards still have too much friction

Let's face it: Many dashboards are still cluttered with (often unnecessary) information, showing information that users "might be interested in" but not the information they're actually looking for. This leads to a poor user experience and to many dashboards being left unused. LLMs have the potential to minimize the information displayed on dashboards and instead focus on making it as easy as possible for users to retrieve the information they want using natural language questions or queries.

5. Many business users still lack data literacy

Many business users still lack the technical skills to create reports and dashboards on their own. LLMs would provide the ability to easily interact with data and ask questions without having to have such a technical background. In this way, even users who previously didn't have access to corporate data can use it in their daily work and thus accelerate the process of data democratization.

So if LLMs will change the way we do BI, then how should you prepare?

5 steps to prepare for LLMs in analytics

Here are my top 5 prescriptions to prepare you for applying LLMs in business intelligence and analytics:

1. Experiment - often

Learn by doing. There is little literature or online courses on LLMs - and even when there are, they tend to go out of date quickly. The best way to familiarize yourself with this new technology is to get hands-on. Experiment with use cases and prototypes, create proof of concepts, and simply try to test the technology's potential in your industry. Don't set expectations too high at the beginning. The goal should be to understand how the technology works and test different scenarios and use cases to make sure it aligns with your company's goals. Review the technology at least once a year, because things evolve very quickly.

2. Collect unstructured data

Start collecting and storing unstructured text data: To get the most out of LLMs, you'll need to collect and store large amounts of unstructured text data from a variety of sources. This includes data from the web, customer service emails, social media posts, and other forms of unstructured data your organization has access to. This data can be used to fine-tune LLMs and gain valuable insights.

3. Modernize your data infrastructure

Invest in the necessary infrastructure and technology: processing unstructured data with LLMs still places high demands on your existing hardware and software infrastructure, as well as the expertise to use it effectively (e.g., GPUs). This may require investment in new technologies or training staff. Cloud computing is an increasingly important part of the modern data infrastructure and can be a key component if organizations are to take advantage of LLMs.

4. Build partnerships

Explore partnerships and collaborations: You may want to seek partnerships and collaborations with other organizations that have expertise in LLMs and natural language processing. This can help your organization gain access to the necessary technology and expertise to take advantage of LLMs.

5. Invest in your people

Giving your employees the opportunity to learn about these new and emerging technologies is an important step in your company's journey toward more efficient analytics processes. In addition to attending conferences, courses and seminars, you can also set up free time, hackathons, or experimentation days so that your employees have the opportunity to learn about new technologies alongside their daily work. In this way, you can ensure that your employees are well equipped to take advantage of the opportunities these new technologies offer.

Conclusion

Overall, organizations should take a proactive approach to prepare for the growing importance of LLMs for analytics and business intelligence.

By exploring potential use cases, collecting and storing unstructured text data, investing in the necessary infrastructure and technology, and exploring partnerships and collaborations, companies can position themselves to take advantage of the benefits of LLMs in the years ahead.

The opportunities LLMs present are too great to miss - so start preparing now!

If you're interested in more data predictions for 2023, keep an eye out for the DBP INSIGHTS: 2023 DATA AND ANALYTICS PREDICTIONS AND PRESCRIPTIONS report where this prediction is part of.

Follow Prashanth H Southekal on LinkedIn so you don't miss it when it's released!

In the next issue I'll also share a link to the full report.

Speaking of the next issue.

AI4BI.Rocks will take a break next week. 😴

Enjoy the holidays and have a happy festive season!

Thank you again for all your support in 2022 and looking forward to an exciting AI year ahead!

Best,

Tobias

AI-Powered Business Intelligence Book Cover

If you liked this content then check out my book AI-Powered Business Intelligence (O’Reilly). Get your copy here: https://www.aipoweredbi.com