Why 2025 will be the year of the AI ​​orchestra

In the tech world, we like to celebrate milestones like the year (insert milestone here). The past year (2024) was a year of wider experimentation in AI and, of course, agent use cases.

As 2025 unfolds, VentureBeat spoke with industry analysts and IT decision makers to see what the year might bring. For many, 2025 will be the year of producers, when all pilot programs, tests and new AI use cases come together with something that looks like a return on investment.

In addition, the VentureBeat experts spoke to see 2025 as the year that AI orchestration will play a bigger role in the enterprise. Organizations intend to make managing AI applications and agents much simpler.

Here are some of the topics we expect to see in 2025.

More practice

Swami Sivasubramanian, VP of AI and data at AWShe said that 2025 will be the year of productivity, as executives will start to care more about the costs of using AI. Productivity assurance becomes essential, and this starts with understanding how multiple agents, both within internal workflows and those that touch other services, can do better.

“In an agent's worldworkflows are going to be repetitive, and you start asking about error and how to achieve fivefold productivity,” he said.

Palantir chief architect Akshay Krishnaswamy agreed that decision-makers, especially those outside the technology group, are starting to get angry about seeing the impact of the This AI investments on their businesses.

“It's right that people are upset about more sandboxing, because it's behind the whole data and analytics journey of the last 10 years, where people also did a ton of experiments,” Krishnaswamy said. “If you're an operator, you're like, 'this has to be the year I start seeing ROI, right?'”

An explosion of orchestration frameworks

Going into 2025, there is a greater need to create an infrastructure to manage multiple AI agents and applications.

Chris Jangareddy, who is the managing director of Deloitteto VentureBeat that next year will be very exciting. Rivals will face him LangChain and other AI companies that want to offer their own orchestration platforms.

“There are a lot of tools coming up with LangChain, and we're going to see more new players coming up,” Jangareddy said. “Even before organizations can think about multi-producers, they're already thinking of an orchestra so that everyone raises that level.”

Many AI developers turned to LangChain to start building a traffic system for AI applications. But LangChain isn't always the best solution for some companies, and that's where some new options come in.Magnetic at Microsoft or Index of lama But for 2025, expect to see more new options for enterprises.

“Orchestration frameworks are still very experimental, with LangChain and Magentic, so you can't be the frontrunner for just one,” said PwC global commercial technology and innovation officer Matt Wood. “Tools in this space are still early, and it's not going to grow. “

Better agents and more integration

AI agents became the biggest trend for enterprises in 2024. As organizations prepare to integrate multiple agents into their workflow, the ability for agents to cross from one system to another system becomes more apparent. This is especially true when companies are looking to demonstrate the full value of their agents to executives and employees.

Platforms like Basics of AWSand even Slack, offering links to other agents from Salesforce's Task force or service now, making it easier to move context from one platform to another. However, understanding how to support these unions and teaches orchestrator agents identifying internal and external agents will be an important task.

When agent workflows become more complex, the recent crop of more powerful reasoning models, such as OpenAI's recent name 03 or Google Gemini 2.0orchestrator agents could make it more powerful.

However, all this will be in vain if enterprises do not get their employees to use new AI tools in 2025.

Don Vu, chief data and analytics officer at New York Lifetold VentureBeat that the last mile problem of workers often choosing more manual methods over AI will continue for the next year.

“The last mile problem is something that we all have to keep on our toes in 2024, and understand that change management, business process reorganization are things that are perhaps not as generic as building an agent can do these amazing things,” said Vu. “It's harder to change human behavior than using an app.” “