Agent orchestration layer should be independent, says DATE CEO

(Technovecuvenors/ Shutterstock)

We are still in the AI ​​agency, but several things become clear: customers will not tolerate the lock or data or AI level, and the coordination of agents with uniform administration will be necessary. For the CEO and co -founder Florian Douetteau, an emerging layer of data orchestration that the company builds.

Sentle was founded in 2013, dates trying to make it easier for users to create data products. In the first days, the company gathered around the cause of science about data and advanced analysis. Recently, the organizational principle of generative AI and AI AI.

During the recent Snowflake Summit in San Francisco, Bigdatawire Dohnal dataics to get the company’s activities update. As Douetteau explained, the pace of innovation in the AI ​​world is at the same time exciting and potentially lucrative as new needs appear.

Three large public cloud platforms and other data platform providers such as Snowflake, Databricks, Salesforce, Servicenow and Workday can allow customers to build AI agents running on their platforms and work with customers that Douetteau Notes there. But AI agents developed by these data platform providers will not require them to be able to work outdoors where dataics come.

Co -founder of the Dataik and CEO Florian Douetteau

“We see this gap on the market,” Douetteau said. “Theoretically, you can build agents on data platforms to ask the data platforms themselves, which is great. However, many interests combine everything that combines everything togger and more steps, and do complex things.”

Most data infrastructure that businesses need to create AI agent systems already exist or can be easily shaken in the cloud. This magazine is similar to the operating system for AI and includes huge storage of objects, integrate data in real time, database of application levels such as Postgres, and vector database for power load RAG, not to mention computing and network requirements.

“If you want to create a value in the company, you also need to allow people in the field to pave things together to create the necessary basic artifacts for applications and agents,” he said. “These things are largely necessary to fill the layer between the axes’ core and the data.”

However, AI AI requires another layer that is not easily available on AWS, Azure or Google Cloud or other retailers, Douetteau said. In addition to the development of AII, you must be able to manage life, which means testing, deploying, monitoring and reporting agents, he said.

Ideally, the development, testing, deployment, monitoring and reports on AI models are not something that developers would have to connect together, which would create unnecessary pain for customers, Douetteau said. Management, safety and auditability are improving if this particular layer of the tank is standardized by one set of tools, he said.

“Ideally, you want it to be firmly integrated instructions that it has five tools – to copy all data together, one to define business tools in a collaborating way, one to evaluate an agent proposing an agent who manages security of your tools and these frames,” said Douetteau.

Douetteau assumes that businesses are building AI agents to manage a number of tasks, whether they are insurance demands or optimize the warehouse. Some agents will be trained to handle very specific tasks, while other agents will operate more as coordinators. The command will look like a tree, with branches and agents AI (or leaves) at the end.

It will be difficult to obtain these complex agency environments for smooth functioning, but it will be easier with an orchestration application that has proven technology in the core.

“(That’s why you need an orchestration,” Douetteau said. “Even if you build complex agents, there is usually one part in the middle at the end of the day, which is prepared a very good rules -based system, or an old old prediction model because you need where you have some guanus about how they work.

Having an orchestration layer that is naturally independent, allows customers to build agents working with any data, no matter where they are located.

“Our vision is that in the end we want to be easy to use AI desk, a place where people who are analysts in the field can go to create a new data product without having to call scientists,” says Douetteau.

Related items:

DataIic triggers a real -time Genai solution

DataIic offers $ 400 million in search of AI democratization

Create your own AI