What’s Ahead in Generative AI in 2025? (Part One)
Predictions for AI in the coming year.
- By James G. Kobielus
- December 18, 2024
The generative artificial intelligence (AI) boom will continue unabated over the next several years. As 2025 approaches, I predict the following key trends in generative AI.
Prediction #1: Expanding Generative AI Usage
Adoption of generative AI will accelerate across business and consumer markets in 2025 and beyond. Already, generative AI features are becoming a key differentiator in enterprise applications in the form of conversational user interfaces, embedded digital assistants, natural-language chatbots, task-oriented copilots, and other large language model (LLM)-powered productivity augmentations.
TDWI research shows that productization of generative AI is starting to gain traction in enterprises. A recent TDWI survey on enterprise AI readiness showed that 30% of organizations have already deployed the technology or have expressed a significant interest in adopting it in the coming year. By the same token, 46% are discussing how generative AI could be useful in their businesses.
Prediction #2: More Transformative Experiences
Generative AI will transform the experience of more devices, applications, and services in 2025. Already, voice chatbots powered by generative AI are becoming a popular choice for the self-service experience of every app, including no-code business analytics tools.
TDWI research shows that generative AI is rapidly changing how enterprises get work done. A recent TDWI survey on data and analytics trends showed that 37% of organizations have already built or are planning to deploy generative AI chatbots for customer support applications, 28% for marketing content generation, 26% for onboarding new employees, and for many other use cases. Over the next several years, we predict that generative AI will democratize many professions as knowledge workers adopt tools that enable them to single-handedly produce sophisticated deliverables that once required highly skilled teams and expensive, time-consuming production processes.
TDWI research also shows that the generative AI application experience will grow more multimodal. In 2025 and beyond, multimodal LLMs will proliferate as developers create more generative AI apps that work with text, image, video, audio, and other data modalities. In the recent TDWI survey on achieving success with modern analytics, 46% of enterprises report either using generative AI text prompting now to generate images and other modalities or plan to do so over the next few years.
Prediction #3: More Sophisticated Platforms and Tools
Generative AI will be a principal driver in the evolution of enterprise data foundations in 2025.
Demand for multimodal applications of generative AI will drive evolution of enterprise data lakes, data pipelines, data governance, and much more. Likewise, adoption of generative AI will bring such platforms and tools as vector databases, prompt engineering, and retrieval-augmented generation into the enterprise mainstream.
In the coming year, prompt engineering will continue its rapid maturation into a substantial body of proven practices for eliciting the correct output from LLMs and other foundation models.
Within generative AI development tool sets, embedding libraries will become an essential component for developers to build increasingly sophisticated similarity searches that span a diverse range of data modalities. The recent TDWI survey on enterprise AI readiness shows that 28% of organizations already use or are deploying vector databases to store vector embeddings for use with AI models, while 32% plan to adopt those databases in the next few years.
In addition, generative AI developers in 2025 will have access to a growing range of tools for no-code development of “agentic” applications that provide autonomous LLM-driven copilot, chatbot, and other functionality and that can be orchestrated over more complex process environments. Agentic workflows, leveraging a new phenomenon known as a “large action model,” will be a growing focus of intelligent process automation.
Prediction #4: More Diverse Models and Data
Developers will have access in 2025 to a growing range of sophisticated models and data for building, training, and optimizing generative AI applications—including both commercial and open-source models.
The recent TDWI survey on data and analytics trends showed that around 25% of enterprises are experimenting with private or public generative AI models, while 17% are building generative AI apps that use company data with pretrained models. The survey also showed that 41% of enterprises are exploring how to use generative AI to build apps that use private data.
Online cloud marketplaces will become a favorite source of third-party training data for more enterprise AI developers. I also predict that enterprises will increasingly rely on synthetic data generation as an alternative for training generative AI in use cases where authentic source data is either too costly, restricted, or unavailable for their needs.
Read more predictions about AI governance and ROI in Part Two of this article.
About the Author
James Kobielus is a veteran industry analyst, consultant, author, speaker, and blogger in analytics and data management. He was recently the senior director of research for data management at TDWI, where he focused on data management, artificial intelligence, and cloud computing. Previously, Kobielus held positions at Futurum Research, SiliconANGLEWikibon, Forrester Research, Current Analysis, and the Burton Group. He has also served as senior program director, product marketing for big data analytics for IBM, where he was both a subject matter expert and a strategist on thought leadership and content marketing programs targeted at the data science community. You can reach him by email ([email protected]), on X (@jameskobielus), and on LinkedIn (https://www.linkedin.com/in/jameskobielus/).