TechVoices

VIDEO PODCAST

Incorta's Osama Elkady: Building the Data Foundation for AI Agents

James Maguire

September 7, 2026

As enterprises race to deploy AI agents, Incorta Co-Founder and CEO Osama Elkady argues that the real challenge is not simply choosing a more powerful AI model—it's building the data foundation that allows agents to work reliably.

In this TechVoices interview, Elkady explains why agents need carefully structured business context rather than unrestricted access to raw corporate data. He also details why deterministic systems remain essential for financial and operational calculations, and why data platforms are evolving beyond analytics into environments for building and governing AI-powered applications.

Looking ahead, he predicts that increasingly capable AI development tools could allow enterprises to replace portions of traditional SaaS and even ERP systems, while large-scale agent deployments will require real-time access to billions of records with sub-second response times.

Core Takeaways

Context Is Critical for AI Agents

AI agents cannot simply consume vast amounts of enterprise data and reliably determine what matters. Elkady argues that companies need a semantic or business context layer that explains tables, columns, relationships, business processes, and other information so models can formulate better questions and deliver more accurate answers.

AI Still Needs Deterministic Systems

Large language models are designed for prediction and reasoning, not precise financial calculations. For tasks such as revenue recognition, inventory, or financial close, Elkady says AI should interpret and formulate the request while a deterministic data engine performs the actual calculation against trusted enterprise data.

Data Platforms Are Becoming Application Platforms

Incorta sees enterprise data platforms moving beyond traditional BI and analytics toward environments where users can build, deploy, and govern AI-powered business applications. Elkady argues that this could allow companies to build highly customized applications rather than continually buying broad SaaS products and using only a fraction of their functionality.

Enterprise Agents Need Mission-Critical Data Infrastructure

Large populations of autonomous agents will require real-time data, access to billions of records, and sub-second answers. Without those capabilities, Elkady says agents will operate on incomplete or outdated information and may ultimately be slower than the humans they are intended to augment.

KEY QUOTES

Why Enterprise AI Needs Context

“Companies started building a layer on top of the data, and they started calling it context, and they try to define as much information in that context to help the models understand the question better by understanding the data better. So the more context you have around the columns and the tables and the views—and even giving more documents around that as well about how the business answers the questions—the LLM will understand from there.”

"If you don't have that context or the semantic, you have to share the entire data set with the model. The model will read through the data and understand the context of the data, but this is a huge data breach risk, and the cost is prohibitive.”

Why LLMs Should Not Do the Math

“Models were not built to do calculations for you. Models were built to do some prediction, some assumption, get you to a certain point. But after that, you need an engine that you can trust to be able to do these very complex calculations.”

“The engine has to be built around the model of that company. Every company does revenue recognition differently than other companies. The way you do revenue recognition from direct business is different than channels, different than partners, and so on. You can train models to understand exactly how your company runs, but you cannot depend on large models with knowledge of the whole world to be able to do that for you.”

From Analytics to Building Applications

"Data platforms should not be just analytics anymore. A data platform should be an operational data platform, which means customers don't just use our platform to get data from external sources. Customers can use our platform to build on the platform. What Incorta realized is customers are not looking just for analyzing—they're looking for building.”

“They want to build solutions that they had to pay for. The deployment is the missing piece: where the data will be stored, how the data will be managed, and how you govern the entire application and the entire lifecycle of the application. Right now so many business users and so many IT people are building a lot, and the company can end up with tens of thousands of applications running without controlling cost, without controlling data security, without any control.”

The Data Foundation for Armies of AI Agents

“These agents will require three things. They require real-time data, they will require access to billions of records, and they require answers in sub-seconds for these agents to be efficient and working. Otherwise, these would be completely useless agents, because if the data is not real time, they are working on data that's completely not relevant.”

“If they don't have access to billions of records of data, they cannot get you the right answer because they have only a subset of the data. And if the answers are not in sub-second and every agent with every question has to wait 10 minutes to get the answer, these agents will be completely useless. A person will say, ‘I can do the job much faster than that.’”

ABOUT THE AUTHOR

James Maguire

Executive Director

An award-winning journalist, James has held top editorial roles in several leading technology publications, covering enterprise tech trends in cloud computing, AI, data analytics, cybersecurity and more. He regularly communicates with industry analysts and experts and has interviewed hundreds of technology executives. James is the Executive Director of TechVoices.

Cloudera's Venkat Rajaji: advantages of open source AI

Snowflake's Christian Kleinerman: AI and the speed of data access

David Linthicum: LLM trends

Extreme Networks' Nabil Bukhari: how AI democratizes technology

Komprise's Krishna Subramanian: an AI model for every task

Ahead of the curve, every week.

Get in touch with us.

Ahead of the curve, 
every week.