The next wave of AI has arrived: Multiagent AI systems are poised to transform the enterprise as we know it by automating complex tasks, across various business departments, with minimal human intervention. The technology is still in its relatively early days, but as it continues to develop, collaboration between AI agents will deepen, and their ability to make intelligent, autonomous decisions will unlock a level of efficiency that has never before been possible.
Given the promise of the technology, it’s no surprise that the global agentic AI market is forecast to skyrocketfrom $7.3 billion this year to $41.3 billion in 2030. Furthermore, Nvidia CEO Jensen Huang notably stated that AI agents represent a “multitrillion dollar opportunity” in his keynote at CES this year.
What will this look like in the enterprise? Gartner predicts that 33 percent of enterprise software applications will include agentic AI by 2028, automating 15 percent of day-to-day work decisions. And over the next decade, AI agents could automate up to 70 percent of office work tasks.
Enterprises need to determine their strategy now in order to stay competitive in this rapidly evolving space. Most organizations are still determining how to effectively operationalize single AI agents, let alone multi-agent systems. Taking a measured, “crawl, walk, run” approach will be critical for success. Trying to deploy too many agents too quickly can result in system failures, loss of trust, and inefficiencies that set the organization back.
Here are five key considerations that enterprises need to be thinking about as they begin to build multiagent AI systems:
Determine Precise Use Cases
Rather than building broad, general-purpose agents, organizations should focus on creating purpose-built agents to address well-defined business problems. Zeroing in on targeted use cases gives organizations the ability to measure, optimize, and iterate on agents to ensure their efficacy before they introduce other agents into the mix. If organizations try to build generalized agents too early, they risk creating unnecessary complexity and making outcomes difficult to measure.
For example, an online retailer might start with a purpose-built agent for returns and exchanges, and track its resolution times and customer satisfaction before enabling collaboration with AI agents across other customer service functions.
Reimagine Business Processes
One of the most notable ways multiagent systems will disrupt the enterprise is by forcing companies to rethink and redesign their processes. Organizations will need to account for where to integrate AI agents into their workflows, which agents will work together, develop triggers for automation, and—importantly—determine where human input fits into the loop.
For example, in procurement, a company might start using AI agents from different departments of the business to generate RFPs. Accordingly, they’ll need to update their processes to include a human in the loop to validate the content of the RFPs, as well as set parameters around how feedback is used to continuously refine the process.
In addition, this change management could be incremental. For example, in customer support, an AI agent could initially help customer support professionals answer queries quickly and obtain modest productivity improvements.
Reskill Talent as Needed
Although multiagent systems mark a significant shift in enterprise operations, most traditional software engineers are already well-equipped to build them. This is due to the fact that agents are essentially just another piece of software built on newer frameworks and patterns; they just so happen to use large language models (LLMs) to make decisions instead of bespoke instructions. This means organizations are likely to augment their existing talent (upon reskilling) with some experts to introduce software engineering teams to new design patterns and ways of coding that are tailored to multiagent AI systems.
A Top-Down Approach is Crucial for Success
All successful AI initiatives are driven from the top down, and multiagent systems are no exception. Building these systems is a complex, delicate process that requires buy-in and input from the C-suite, in addition to legal, InfoSec, HR, product, and finance teams.
A centralized, top-down approach makes coordinating the investments, partnerships, risk management, governance, and reskilling necessary to build these systems significantly faster and more streamlined.
Have the Right Infrastructure in Place
The right infrastructure is critical for enabling multiagent systems and realizing the full value of the technology. Increasingly, AI workloads will need to run where business decisions are made, not necessarily where data is stored. Enterprises need to invest in hybrid AI infrastructure that spans the cloud (including multiple clouds) as well as on-premises environments. Since agents will be using a company’s private data to gain insights and make decisions, they need infrastructure close to where that data resides.
The reason for this is simply because it’s far more efficient to move the AI to the data than the other way around—and every enterprise will have a different approach depending on their data architecture. Hybrid infrastructure gives enterprises the flexibility, resiliency, and cost-management capabilities they need for multi-agent systems to transform their business.
The race to operationalize multiagent AI systems is on, but enterprises must understand that this process will be a marathon, not a sprint. However, they do need to determine their multi-agent strategy now, and invest in the infrastructure required to support it so agents can operate intelligently, securely, and at scale. Without a clear plan and the right foundation, the promise of multiagent systems will remain just that: a promise.
