At MIT EmTech AI, 2026 is called the year for enterprise agentic AI, with companies deploying agents that autonomously handle marketing, service desks and background workflows.
The pace of enterprise agentic AI deployment is picking up rapidly, and 2026 is shaping up as the year agentic AI really goes to work.
That's the message delivered at the MIT Technology Review's EmTech AI conference, which was held April 21-23 in Cambridge, Mass.
Enterprise agentic AI is moving from a period of integration and experimentation into one of execution, where agents are now starting to perform real work.
This is the year enterprises need to get serious about deploying agents, if they haven't yet, Andrew Bialecki, co-founder and co-CEO at Klaviyo, said in a conference session.
Klaviyo is a Boston-based firm that provides an AI platform for marketing services.
Business people have two main questions around the deployment of AI, Bialecki said. They want to know the best way to market their products and services to grow the business and deliver a better consumer experience.
This will happen through agents, as every business will start to deploy an agent, he said. The agent will be infinitely scalable and able to handle thousands of different tasks and will know all the minutiae on how a company's processes work and how people understand them.
Marketing an agentic AI hot spot
"This is going to be the year like when everybody discovered Claude and ChatGPT last year, but this year everybody will go to a website and expect a similar experience, but one that's trained on that business's information," Bialecki said.
Marketing is a prime area for putting agentic AI to work in the enterprise, as the customer is at the center of all businesses and needs products and services that match what they are looking for, he said. Klaviyo's AI-based marketing platform draws on market data from over 200,000 businesses, providing insights into what consumers are thinking across different geographies and product categories.
"Our AI infused with this data feed can basically see what's happened across the world," Bialecki said. "We can take your ideas or ideas that AI thinks up on its own and then tune them so that they fit what a customer is looking for that day."
Agentic AI can also invert the traditional marketing practice of connecting outwardly to customers through messaging and advertising, he said. Now there's a digital experience where customers are coming to businesses with questions, and it's about how to match up with them.
"This is extending what businesses traditionally thought of as customer service and is driving incremental engagement or sales revenue to that business," Bialecki said.
Traditional structures like customer service, operations and marketing are siloed and work differently, but this doesn't make sense from a consumer standpoint, he said.
"But now with AI, those walls are falling down. You have one agent that's helpful to the customer and another that's in there thinking about what are the things the customer might be interested in," Bialecki said.
Reinvention required
In a session, Kellie Romack, chief digital information officer at ServiceNow explained how agentic AI has infused the company's operations.
ServiceNow decided to deploy AI internally by building agents to handle service desk operations, Romack said. The service desk AI optimization improved service requests from first touch to resolution by 90%, outpacing the project's original goal of 85%, which Romack termed "audacious."
However, in an age of increased anxiety around AI-related job losses, this was done without reducing the head count in the service desk team, she said.
ServiceNow moved 85% of service desk employees to higher-level jobs, and the rest were installed as managers of the AI service agents, where they can resolve cases that the agents can't, according to Romack.
Enterprise agentic AI can only be successful if it's fully understood, she said.
"People fear what they don't understand, so part of my job for our employees, customers and partners is to move AI from a black box to a glass box," Romack said. "You need to understand it, be informed by it and be able to take actions and get outcomes with it. It's not magic, it's technology, it's processes all put together."
But to achieve results, enterprises need to rethink and reinvent processes, not just stick AI on top of existing processes, she said.
For example, ServiceNow's 9000 sales employees have a compelling need to understand the status of their compensation plans at any time, according to Romack. However, the finance team was overwhelmed with constant requests about the plans, and it would take four days to get the information.
We talked about it as humans, reinvented the process, put agentic AI to work and moved a four-day process to eight seconds. That's reinventing work.
Kellie RomackChief Digital Information Officer, ServiceNow
"The finance team said they don't want to deal with more tickets, the sales team said they need faster answers," she said. "We talked about it as humans, reinvented the process, put agentic AI to work and moved a four-day process to eight seconds. That's reinventing work."
Background effects
There are several types of agents – user-driven, such as ChatGPT; more vertically specific ones, like task-specific agents; and background agents for processes that run in an organization's background, said Ash Edwards, head of forward-deployed research engineering at Poolside in a conference session.
Poolside is an AI research lab and developer of specialized AI foundation models and agentic systems based in London.
"The background stuff is most interesting, as companies become more and more agentic, more and more is happening in the background," Edwards said.
Ash Edwards of Poolside speaks at the MIT Technology Review's EmTech AI conference.
Bug fixes and reporting in software engineering is one prime function that background agents are well-suited to do, because tasks like bug fixes aren't the main part of the job, it's just something that the engineer has to do, he said. Handing over this type of tedious and repetitive task can help free up developers to be more productive and creative.
"It's never been a more fun time to be a software engineer because you have an idea and before you might have had five different directions and you had to pick one to do," Edwards said. "Now you can literally try all five and get much more exploration, which means you can be much more creative."
Handing background work over to agents also makes sense for other enterprise teams, he said.
"If you think about sales organizations, so much time is spent on work that's not core to what makes a [sales] person good at their job or is interesting, it's just background work that can and should be automated," Edwards said.
Jim O'Donnell is a news director for TechTarget, where he covers IT strategy and enterprise ESG.
