The way generative AI (GenAI) has fundamentally altered our society is as undeniable as it is widespread. Outside of adjusting to the ways GenAI has changed the way we interact with the world around us, we must also contend with its rapid evolution. While its latest advancements—from AI agents to hyper personalization—promise a wealth of opportunities, an assortment of challenges also come in tow.
To help break down the latest and greatest of GenAI, experts joined DBTA’s webinar—in partnership with Enterprise AI World—Top Trends in GenAI for 2025, offering the necessary expertise and strategies for navigating the fast-paced AI era.
Christian Capdeville, director of product marketing, Dataiku, emphasized that in a rapidly changing space, uniting the different components of your enterprise is key to an adaptive AI strategy. The Dataiku Universal AI Platform unites people, technology, data, and governance within a secured, scalable, adaptive framework that future-proofs your technological investments with an “aggressively agnostic” approach.
This is achieved through the Dataiku LLM (large language model) Mesh, which helps companies “accelerate these GenAI use cases and make sure they are future-proofed, [meaning,] as technology changes, they are still able to get the value from the investments that they make today in GenAI,” explained Capdeville.
The Dataiku LLM Mesh enables you to “swap out all the underlying technologies without having to rebuild anything,” said Capdeville, accompanied by a unified way to manage the development, deployment, and monitoring of advanced GenAI use cases alongside other ML models and applications.
Taking a broader look at today’s top challenges with GenAI, Timo Selvaraj, chief product officer, SearchBlox Software, Inc., examined the following as five key obstacles:
- Solution Complexity: Multiple pipelines, too much glue
- Data Ingestion: Multiple formats and data sources (structured and unstructured)
- Lack of LLM Data Privacy: Sharing private data with external services
- Unpredictable Costs: Embedding, inference, and reasoning costs
- Retrieval Accuracy: Wrong content being retrieved
Solving the first three GenAI inhibitors requires a single platform that provides a unified set of information capable of pulling in data from many different sources—such as SearchBlox’s SearchAI platform. Moving from pipelines to a platform, SearchAI accelerates deployment time with ready-to-use connectors for indexing diverse information paired with native security integrations for data privacy and safety.
Regarding retrieval accuracy, Hybrid RAG is a foremost solution for rectifying the accuracy challenges of GenAI, according to Selvaraj. Capable of improving RAG accuracy and reliability across the enterprise, hybrid RAG retrieves and combines both keyword search and semantic search results, surfacing relevant results even when the term is not found within content.
Maarten Van Segbrook, head of applied science, Gretel, pointed to synthetic data as a significant GenAI trend for 2025. Data is the bottleneck to innovation, where lack of access to high-quality training data is a central problem to succeeding with GenAI.
In tackling this challenge, synthetic data becomes more crucial. Synthetic data is created by GenAI models as an alternative to real world data—”it’s like your data, but better,” noted Van Segbrook.
Gretel helps empower enterprises to embrace a data-centric approach to AI with a synthetic data platform purpose-built for AI use cases. Gretel Navigator allows enterprises to create, edit, and synthesize data, depending on the amount and status of starting data you have. Capable of generating a high-quality synthetic version of your data or creating high-quality synthetic data from scratch, Gretel helps boost LLM accuracy, ensure data privacy, develop specialized tools, and accelerate R&D.
Focusing on the trends shaping GenAI excellence, Aymen Ben Azouz, lead sales engineer, enterprise, Fivetran, explained that “Generative AI is driving faster innovation for businesses of all sizes. But it’s not a silver bullet.”
“As with analytics, GenAI is only as good as its data. That makes data readiness your foundational competitive advantage,” Ben Azouz continued.
With this in mind, Ben Azouz looked at 7 data and AI trends for 2025 that make GenAI the most knowledgeable entity within the organization, capable of unlocking new opportunities for innovation. The following are some of the aforementioned shapers of successful GenAI initiatives, according to Fivetran:
- Data lakes and modern tools will define the future of AI.
- Open data lakes will become the industry standard for driving cost efficiency and enabling innovation across industries.
- Knowledge graphs will play a crucial role in extracting valuable insights from unstructured data.
- Strong AI governance strategies will ensure AI models remain accurate and reliable.
With these trends in mind, Fivetran is a core component of the new AI stack, noted Ben Azouz. As a data integration platform, Fivetran helps move data from your data sources into your data lake, data warehouses, and more. Fivetran automates the process of data migration so that enterprises can focus more time on AI innovation rather than the intricacies of the underlying data infrastructure.
This is only a snippet of the full discussion of GenAI trends for 2025. For the full discussion featuring crucial insights, a roundtable discussion, and more, you can view an archived version of the webinar here.