Enterprise adoption of artificial intelligence agents is accelerating, but many organizations struggle to get these systems to perform reliably in real-world business environments. The gap between general AI capabilities and specific business needs is becoming a major obstacle for companies trying to implement agent-based strategies. Impetus Technologies Inc. has built its value proposition on bridging this “context gap,” which is the distance between what large language models know and the unique data structures of an enterprise.
Deepak Khosla, chief growth officer and head of AI at Impetus, explained during a recent interview on theCUBE that the problem lies not in the models themselves, but in how they are integrated into a company’s operations. “We figured out the gap is not the large language models,” Khosla said. “The gap is the context and that’s why we want to fill that gap.”
Related: Lam Research to invest over $3B in new labs
Impetus has developed a methodology called CEDL, or Context Engineering Delivery Lifecycle, to manage this process. This framework treats organizational context as managed infrastructure rather than a loose set of prompts. The approach involves testing and governing the data that agents use to ensure they function correctly within the specific environment of a business. By moving beyond generic prompting, companies can create AI systems that understand their operational rules and data setting.
Leap AI helps companies move AI into production by focusing on infusing agentic systems with software discipline and service flexibility. This suite enables the creation of a knowledge layer and a semantic layer early in the data platform modernization process. By doing this upfront, Impetus brings enterprise knowledge to the agents as they are deployed. The company builds knowledge graphs to map relationships between entities and creates an ontology layer of business tools. Memory is also a critical component, as agents need short-term and long-term memory to maintain continuity in their operations. If an agent learns incorrect information, it can permanently affect its future actions, making memory management essential.
The implementation of this framework requires a mix of technical upgrades and data organization. Impetus works to modernize legacy systems and organize semantic meaning to help agents understand their surroundings. Deepak Khosla described the necessity of providing context to these systems using an analogy about hiring. “It’s like a new hire, a new hire in the company, a smart person, and you don’t tell them anything about the company, the smart person is not going to do anything well,” Khosla said. “But if you tell the smart hire about your business, your rules and everything, the [person] will do good.” The Leap AI suite applies this logic to software, making agents smarter by equipping them with the specific context required to execute tasks effectively.
Related: Texas data center raises emission concerns
While the technical framework for context engineering is becoming more defined, the practical application remains challenging for many enterprises. The process requires a significant investment in data governance and infrastructure modernization before an organization can fully realize the potential of its AI agents. Companies that succeed in building these contextual layers may find their AI systems far more capable of handling complex, real-world tasks than those that rely on generic model capabilities alone.
Building this foundational infrastructure is essential for long-term success.