Highlight the importance of the origin of data, control of its quality and reliability so that Generative AI effectively leverages business and is used to its full potential
THE Qlik, a leader in data integration and analysis, announces the top 10 data, analytics and Artificial Intelligence (AI) trends that will guide businesses in 2024. According to Qlik, for Generative AI to effectively leverage businesses to their full potential, the The market needs to focus on the origin of data and controlling its quality and reliability. The results of the analysis also highlight important views on hybrid AI, insights, unstructured data, AI literacy, data engineering, analytics, customization, among other aspects.
“We are right in the middle of an AI boom, with Generative AI promising to usher in a new era of productivity and prosperity,” says Dan Sommer, Senior Director of Market Intelligence at Qlik. However, according to the executive, there are growing concerns that limit this promise, due to the lack of traceability of the origin of the data or quality control. This has allowed faulty data to pollute the reliability of Generative AI results with hallucinations, disorientation or even outright untruths, which represents an exponential danger for companies and society. “We need to find a new model that promotes better and reliable data. Trusted data, combined with analytics and automation, will be the foundation for helping people and organizations make better, more efficient decisions, while powering responsible AI.” 
At 10 main technological trends for Artificial Intelligence for 2024 are:
1 – Hybrid AI will address the Generative AI maturity gap: Current technological discussions focus on Generative AI, which actually has broad potential. However, several other Artificial Intelligence efforts are already underway and bearing fruit. One of them is Machine Learning (ML), which is increasingly democratized. It is a mistake to think that Generative AI will replace everything; that would be a big mistake. AI has matured enough to be put into production and scaled, including in well-established use cases like fraud and churn analysis. Therefore, although Generative AI is consolidating, Machine Learning and other Artificial Intelligence already demonstrate unlimited potential and can even be used to address this maturity gap in Generative AI.
2 – Generative AI for insights – Improving the data consumer experience: It is necessary to think about how AI is truly empowering less technical users. In Qlik's view, what less technical information professionals are looking for is a quick response to their needs, as they do not have the time, desire or skills to perform an in-depth analysis. This user base appreciates automatically generated visualizations and insights, enhanced with natural language explanations. Even better if this happens directly on the systems they already operate on. The market must pay attention to this to generate more productive and satisfying experiences for AI and data users.
3 – The era of unstructured data is now: Most of the world's data is unstructured (80%, according to Forester), that is, it is not well organized into rows and columns for analysis. An example of this are emails and documents on company intranets. Many have tried and failed to analyze unstructured data, but with new metadata and semantic techniques it is now possible to unlock it. By utilizing knowledge graphs and vector databases, complemented with RAG (Retrieval, Augmentation, Generation), the opportunities to reliably combine structured and unstructured data are endless. Combined with a response management layer, it is possible to reuse verified and reliable questions and answers, allowing you to analyze companies' entire data assets and use private LLMs (Large Language Models) created internally through data analysis.
4- From Business Intelligence (BI) to Artificial Intelligence and vice versa, business analysis is changing: New ways of interacting with data are emerging at rapid speed. Now you can drag a file into a simple chat interface and start chatting with it, which generates queries and code, helps you build content and speed up automated processes. Individuals can start the analytical journey on these Generative AI tools for simple data visualization and business projections. This is Business Intelligence coming to AI. As a next step, users may want to leverage enterprise-grade tools for deeper analysis, bringing the benefits of Generative AI to their tools. This is AI coming to BI. The market will switch between these two modes – enabled by embedability, connectivity and APIs – to derive maximum benefits from each platform.
5 – Where data comes from matters: Understanding the DNA of data: If you don't know where the data comes from, you can't trust it. Data quality and lineage have become non-negotiable in an AI world, especially for training AI models. The need for identifiable and understood data sources is essential in public LLMs, which do not currently track them. Without this knowledge, it is difficult for the best Generative AI models to differentiate fact from fiction. This can lead to hallucinations, false facts and deepfakes. For companies, relying on results like these can have serious consequences. A mechanism is needed to clearly label and flag data, using provenance and encryption techniques with techniques that have not yet been invented to create the equivalent of a “DNA test for the data.”
6 – New Developers Demand AI Literacy: In a short time, there was an evolution of low code as the new dominant programming language. Simplified coding makes more advanced tasks like app creation easier. This will generate an explosion of applications created by the 'everyday developer', who is not as specialized. While this can lead to a new wave of innovation, it can also lead to chaos in governance and too many applications. As this process puts very strong powers in the hands of the many, organizations will need to educate their workforce about the benefits and pitfalls of Generative AI. If the last five years were about teaching teams data literacy, now we need to shift to AI literacy. Furthermore, managing the application lifecycle and promoting data and apps correctly will take on new importance.
7 – Data engineering, analytics and data science are merging: New data platforms, combined with evolving data fabrics, will “consumerize” data engineering for a new generation of users, especially if they are enhanced with powerful AI, automation, and data science. This will enable business analysts to perform data management and preparation tasks. They will also be able to apply advanced statistical models to the data and tools they work with every day, without having to export it to an advanced environment. Making tasks easier and merging the functions and capabilities of data engineering, data science, and analytics will enable organizations to solve more difficult problems. Adding more connection between previously siled functions will help companies migrate data and achieve better results.
8 – Automation and AI create a virtuous cycle: Until now Large Language Models and Generative AI have been used to support reasoning and analysis. But there are now several efforts underway to support effective execution, including an approach that involves the synergy of reasoning and actions, to create agents that will be able to plan and execute complex initiatives, bringing AI to automation. This requires transformed data in real time and in the right place. We will start to see new ways in companies to use Generative AI with application automation, such as using sentiment analysis to automate and generate different responses depending on mood. Generative AI, linked to automation, will mean less manual work for people to connect and create workflows and instead take on the role of decision curators.
9 – The last mile of AI customization becomes business critical – The first applications of Generative AI are extremely scalable, yet generic, projects that can use Large Language Models in a consumer-focused (B2C) context. Over time, we will see more AI customized for the enterprise (B2B) market, using private applications whose base is common, but with layers of customization that better serve the “long tail”. We will see that with less effort and hours of consulting, sophisticated applications for a specific industry or problem will emerge. Proprietary organizational data will be a valuable raw material in this context and “solution fabrics” will emerge in which domain-specific data and applications can be shared and traded. However, the question of what AI will be the basis for building this remains unanswered.
10 – Data as a product that can be commercialized – Approaches to harmonizing data like data fabrics and data meshes have gone from hype to reality over the past year due to AI and technological advances. A key component of these approaches is “data as a product,” a concept that applies product management principles to data, addressing what problems need to be solved, what the data will be used for, and by whom. Data as a product is evolving to become the basis of consumability for all forms of analytics and AI. The concept of product data indicates that it can be displayed in a catalog, used for various purposes and even evolve into a tradable good. The goal is to monetize data as a product outside of organizations. Companies will now use their own data to further train ChatGPT models, which can then be monetized. In the future, similar exchanges will serve as sources that can track sanctioned data and pay for access, as the music industry has done with streaming services. The more the data product is used, the more valuable it is.
EN 
PT 
