“While a lot of initial generative AI work has been about asking questions to models and summarizing the data, the agentic push is seeing more organizations move towards how LLMs can get actions done,” explains Ajay Nair, General Manager of Platform at Elastic.
As organizations ramp up their generative AI capabilities, a frequent challenge arises: how to maximize the effectiveness of AI, especially Large Language Models (LLMs). Ensuring the outcomes generated from these AI systems are not only accurate but also relevant to the business context is critical.
Elastic is positioning itself to help businesses tackle this challenge. In a recent conversation with CRN Asia, Ajay Nair emphasized that search plays a pivotal role in powering AI, asserting that generative AI signifies a transformative shift in how organizations harness information to extract value. “We believe search is the heart of AI, and we excel at being that particular one,” he stated.
Nair further elaborated on the evolving understanding of ‘speed’ within organizations, as they adapt to the rapid consumption, processing, and actioning of information. According to him, technology must evolve to meet this pace. A significant concern for customers is ensuring that their systems and architectures offer flexibility—not just in terms of choosing models but also in how these models can connect with relevant data.
From Elastic’s standpoint, the journey towards generative AI is more than a mere transition; it embodies a philosophical shift. Organizations are moving from generating answers to facilitating action. “There’s a big push among companies today to explore how LLMs can enable actionable outcomes,” Nair noted. This transition demands robust guardrails to ensure LLMs can effectively reason and act in a predictable manner.
The AI Journey
For Nair, Elastic’s core philosophy has always revolved around building systems that facilitate meaningful engagement with data. This philosophy aligns with the unique maturity journey of their customers. Many businesses are grappling with unstructured data volumes and are seeking coherent methods to access, connect, and utilize this data effectively. “Often, operational data resides in Elastic, and customers look for integration with software applications like Workday or Salesforce,” he mentioned.
Key questions emerge regarding the relevance of the data and how it aligns with organizational needs. Nair stressed the importance of preventing ‘hallucination’—the generation of responses by LLMs that deviate from reality. Observability of LLM actions is becoming a focal point, as organizations recognize the risk of replicating models without context. “Our conversations with clients often explore how they can optimize their workflows using AI effectively,” he remarked.
Customers are becoming increasingly aware of the need to understand their unique business workflows. Many have relied on custom approaches to manage data, often overlooking automation opportunities. “AI is pushing the envelope much further, compelling organizations to reevaluate whether certain processes are indeed essential,” Nair explained. This is particularly evident in areas like customer service, where traditional engagement models are evolving, leading to higher automation levels.
Reflecting on Elastic’s internal use of LLMs, Nair stated that they have realized a 30% improvement in customer support efficiency through automation, showcasing tangible benefits. However, the maturity curve differs significantly among clients. For some, the issue is just identifying the data necessary for effective operations, while others have the data but struggle with translating it into actionable steps.
Knowing What Works Best
Nair identified three pivotal areas organizations need to focus on during their AI journey: trust, flexibility, and actioning data. Firstly, trust hinges on ensuring that LLMs deliver reliable information. This requires broad and secure access to data that maintains high relevance. “Elastic stands out in this realm, with a recent innovation yielding a 120% improvement in recall,” he noted.
Secondly, flexibility and openness are crucial. Nair highlighted that many AI providers offer simplified solutions, but the rapidly evolving AI landscape necessitates avoiding restrictions that could stifle innovation. “Choosing an open ecosystem allows organizations to leverage a diverse range of model and data options,” he advised.
Lastly, the ability to act on data is critical. Many systems falter at retrieval, but organizations need to establish deterministic workflows that connect their data to other systems effectively. “Elastic excels in these dimensions with capabilities like agent builder workflows and inference services,” Nair elaborated.
Data Sovereignty
In discussing how organizations can ensure their AI systems do not expose sensitive information, Nair emphasized the importance of a robust context engineering platform. “This challenge has persisted even before AI. Human operators must only access data they are authorized to see; this principle extends to AI systems as well,” he explained.
Elastic fosters deep privacy controls that govern data access, enabling organizations to set stringent rules around who can access which pieces of information. “We allow organizations to filter out sensitive data programmatically or through traditional methods,” he added.
Organizations must adopt a conscientious strategy when implementing AI, ensuring it operates within a secure framework. “The fundamental step is partnering with a reputable data platform—one that understands and implements data sovereignty principles,” Nair advised. With Elastic’s global reach, businesses can store data in over 50 regions, complying with varying regulations to uphold data privacy standards.