Currently working as Head of Data and Digital for Cerba Research, Martijn Bauters oversees the existing Data and BI team, and leads the company’s digital transformation. Martijn earned a Bachelor’s degree in Wetenschappen-Wiskunde at Sint-Bernardus College, a Master’s in Business Engineering: Marketing Engineering at Ghent University, and a Master’s in Statistical Data Analysis, also at Ghent.
Martijn’s specialisms include statistical data analysis, marketing engineering, and programming, and he’s experienced in sector-specific public speaking. In this post Martijn introduces the concept of interactive insights, and explains how generative AI could transform the way we interact with data analytics.
What opportunities do LLMs and text-to-speech models offer to the data insights industry?
Over the last years I have specialised in leading data projects from both a business and technical point of view while transforming data lab teams into real delivery factories. We have witnessed multiple (r)evolutions within the realm of analytics.
We transitioned from handwritten tables during the early days of professional businesses to Excel spreadsheets with the advent of computers. Following Excel, we embraced Kimball’s principles and moved towards enterprise data warehouses that facilitated data aggregation and manual value creation. In recent times, we have progressed from automated dashboards to self-service analytics, guided by the principles of effective storytelling. Now, a new era of analytics dawns, where generative AI plays a pivotal role in shaping interactive insights.We have been astounded, surprised, and captivated by the music and art pieces generated by generative AI.
The traditional GANs have evolved into more comprehensive concepts such as LLMs and LLaMas, culminating in the release of GPT models to the public towards the end of 2022, with offerings like ChatGPT and similar competitors such as Bard. If AI can create music and art, it can also generate and interpret data from well-designed data platforms, harnessing this information to produce valuable insights.
This potential, combined with new AI applications like Whisper or Meta’s latest text-to-speech model, will revolutionise the way we interact with our insights. Welcome to the era of Interactive Insights.The concept behind interactive insights is not merely connecting our modelled data (the serving layer) to a traditional data visualisation tool, but rather building an interactive layer between the data platform and the end-user.
The latest LLM models can effortlessly interpret well-designed data platforms, extract meaningful data and transform it into understandable and actionable messages. Moreover, these models can comprehend questions posed to them, retrieve relevant data and transform it into information for the requester.
This integration can seamlessly fit into any company’s existing technology landscape, envision a chatbot within your communication tool (Slack,Teams, etc.); or imagine generated information embedded within your dashboards, explaining what you observe and helping to identify anomalies.
When we combine the power of these algorithms with text-to-speech capabilities, we suddenly have the ability to engage in a phone conversation during public transport or while stuck in traffic on our daily commute. Just imagine every decision-maker having a Data Scientist in their pocket, ready for use whenever needed. The amalgamation of LLMs, text-to-speech models, and well-designed data platforms will usher in a paradigm shift in the insights industry and potentially eliminate the necessity for traditional dashboards and data analysts. In the coming months, the industry will embrace interactive insights due to the following advantages:
These benefits will result in cost and resource optimisation, empowering decision-makers to gain faster insights and make better-informed decisions.
Nevertheless, these technologies are still relatively new and prone to errors, as evident in some of the inaccuracies produced by ChatGPT. To overcome these challenges, we need better algorithms and robust enterprise data models that can effectively support interactive insights. It is crucial to avoid falling into a ‘chicken-or-the-egg’ situation, whereby decision-makers act based on imaginary information, inadvertently shaping the future of the business.
In conclusion, the mainstream adoption of interactive insights as the primary method of interpreting information will take some time considering the necessary setup of our data platforms and the ongoing challenges faced by current generative AI platforms. However, we can observe significant shifts within the data visualisation tools landscape, such as PowerBI’s integration with OpenAI in the Azure Fabric, as well as emerging startups like Ficus Analytics venturing into this space to offer interactive insights.
