As companies begin to adopt "context engineering" as a core strategy to enhance the reliability and accuracy of AI systems, this technology is emerging as a new essential element in the AI era. At the AWS re:Invent 2025 conference, Elastic's Chief Product Officer Ken Exner emphasized that large language models must be designed to operate "at the right time, with the right data, within the appropriate scope" to yield reliable results.
Exner stated: "Many companies today face the limitations of mere prompt engineering when implementing agent AI. To successfully build AI applications, it is crucial to continuously provide the correct context for LLMs." He referred to this as "context engineering" and predicted that it will become a core concept in future AI development.
As the independence of AI models in judgment and action continues to increase, Exner pointed out the need to be vigilant about the potential errors or uncertainties that may arise from a lack of context. To this end, various supplementary methods such as retrieval techniques, tool-based reasoning techniques, and memory systems are being introduced. He explained: "LLMs are essentially systems that predict the next word, and this process can only yield more consistent and reliable results when conducted within an appropriate data range."
Elastic has developed the 'Elastic Agent Builder' solution to address these technological challenges. This tool supports the construction of sophisticated agent applications by combining user-customized prompts with data indexing capabilities, and it includes built-in basic conversational agents to help users easily create their own dedicated AI agents.
The criteria for assessing the success or failure of context engineering are also being established. In this process, 'evaluation' and 'observability' play important roles. Exner stated: "Agents now need to be treated as core systems, with an emphasis on performance validation and quality testing. Such validation should operate like unit tests, while also incorporating integrated testing using LLMs as evaluators."
In the quest to ensure the reliability and sustainability of AI systems among enterprises, context engineering is transcending the realm of buzzwords and spreading into substantial technological foundational strategies. Exner envisioned at this conference: "In the coming year, we will hear the term 'context engineering' more frequently, and this field will play a decisive role in AI's progression to the next stage."