Future Of GenAI: Most Apps Will Run On Existing Data Platforms By 2028, Gartner Predicts

Organizations looking to harness the power of Generative AI (GenAI) for business applications are poised to make significant advancements by 2028. According to Gartner Inc., a staggering 80% of GenAI business applications will be developed using existing data management platforms. This strategic move is expected to halve both the complexity and development time of these applications, marking a significant leap towards efficiency and streamlined operations.

Prasad Pore, Senior Director Analyst at Gartner, emphasized the current challenges in GenAI application development. He pointed out that the integration of large language models (LLMs) with an organization's internal data, coupled with the adoption of rapidly evolving technologies, presents a complex landscape. Technologies such as vector search, metadata management, prompt design, and embedding require a unified management approach to avoid prolonged delivery times and unnecessary expenses.

The evolution of data management platforms is crucial for developing GenAI-centric solutions. These platforms must incorporate new capabilities or services that cater to GenAI development. This evolution ensures that organisations are prepared for AI deployment and can successfully implement these advanced applications. As data management platforms transform, they become more conducive to integrating GenAI technologies, thus facilitating smoother and more effective application development.

Retrieval-augmented generation (RAG) is emerging as a pivotal technology for deploying GenAI applications. It offers flexibility in implementation, improved explainability, and better composability with LLMs. RAG works by leveraging data from varied sources, both traditional and non-traditional, to provide context. This enriched context boosts the performance of LLMs in downstream GenAI systems, leading to more reliable and effective applications.

Prasad Pore further elaborated on the importance of combining LLMs with business-owned datasets through the RAG architectural pattern. He noted that most LLMs, which are trained on publicly available data, struggle to address specific business challenges effectively on their own. However, their accuracy greatly improves when coupled with RAG. In this process, semantics, especially metadata, are vital. Data catalogs capturing semantic information can significantly enrich knowledge bases, ensuring the appropriate context and traceability for data used in RAG solutions.

To navigate the complexities of deploying GenAI applications successfully, enterprises are advised to focus on a few key strategies. First, assessing whether current data management platforms can transition into a RAG-as-a-service platform is critical. This approach could replace standalone document/data stores as the primary knowledge source for GenAI applications.

Furthermore, prioritising RAG technologies like vector search, graph, and chunking, whether from existing data management solutions or ecosystem partners, is essential. These technologies offer resilience against technological disruptions and are compatible with organizational data. Lastly, leveraging both technical and operational metadata is crucial for protecting GenAI applications against misuse, privacy breaches, and intellectual property leaks.

As organisations adopt this approach, guided by the recommendations provided, the deployment of GenAI applications is expected to become more streamlined, secure, and successful. This transition not only simplifies the development process but also ensures that GenAI applications are more aligned with business needs and challenges, paving the way for groundbreaking innovations in various industries.