AI agents loom large as organizations pursue generative AI value

Agents come in many forms, many of which respond to prompts humans issue through text or speech. Yet as organizations figure out how generative AI fits into their plans, IT leaders would do well to pay close attention to one emerging category: multiagent systems. 

In such systems, multiple agents execute tasks intended to achieve an overarching goal, such as automating payroll, HR processes, and even software development, based on text, images, audio, and video from large language models (LLMs).

Eighty-two percent of leaders surveyed by Capgemini said they expect to integrate them into their businesses to help automate such tasks as generating anything from emails to software code to analyzing data within the next one to three years. 

How multiagents operate depends on the tasks and goals they’re designed to accomplish. 

All aboard the multiagent train

It might help to think of multiagent systems as conductors operating a train. You’ll have a lead conductor—a “boss” if you will—who doles out tasks to a series of other conductors, or subagents.

A human user might query the lead conductor through a classic user interface, such as an LLM prompt window, thus setting off a chain of events as each subagent handles a different task.

The agents may collaborate with each other, other digital tools, systems, and even humans, tapping into corporate repositories to gain additional organizational knowledge. Importantly, these systems learn from their task history, human feedback, and other inputs to regularly improve their performance as well as adapt to changes in their environment. 

Essentially, they are self-governing and iterative, not unlike human employees. At a time when organizations are seeking to generate value from GenAI, multiagents hold perhaps the most promise for boosting operational productivity.

The value that agents can unlock stems from their potential to automate complex use cases characterized by highly variable inputs and outputs—use cases that have historically been hard to automate. 

Multiagents make automation actionable

McKinsey landed on an excellent example: booking business trips. Think of all the logistical planning and steps to navigate as you secure various travel arrangements, lodging, meals, etc. Sure, some of this has been automated in some capacity, but it still requires a wide variation in inputs and outputs that have historically proven too costly or technically challenging to automate.

Now imagine a business using agents for “actionable automation,” across sales and marketing, HR, IT operations, and other functions. 

An agent could create a net new sales analysis, working with other agents to scan the various sales inputs and outputs for relevant information, draft a document, review it, vet it against corporate standards, and revise it accordingly. McKinsey cites loan underwriting, code modernization, and marketing collateral among other potential knowledge work use cases.

However, multiagent automation needn’t be limited to the digital realm; agents might also manage electrical systems, from elevators to HVAC, controlling temperatures and lighting across zones. Such systems are already highly automated.

These cases are largely conjectural; your use cases will depend on your business needs.

However, automating such work would free human employees to focus on running some of the softer aspects of the business, including collaborating more with colleagues and interacting with customers—tasks that could improve employee and customer promoter scores.

Of course, ensuring digital resiliency remains a challenge with multiagent systems. That is, if one agent fails, will the entire system break down?

This is something the tech industry has seen time and again with robotic process automation, where bots taking their cues from rules-based programming got stuck when a variation presented itself. Truly autonomous agents must correct themselves so they can achieve their goals.

Until then, having a human-in-the-loop who can initiate kill switches or execute rollback capabilities as you begin to experiment with multiagent systems will be critical. 

Preparing your organization for a multiagent future

Ultimately, the key is ensuring that multi-agent systems operate in alignment with organizational goals to achieve the desired business outcomes. As an IT leader, you must be ready to support agentic systems should you and your business stakeholders elect to pursue the option. 

From your front-line coders to your DevOps practitioners and hardware engineers, your organization must be ready to adapt to dynamic change—whether it’s implementing single digital assistants or fleets of autonomous agents.

Distributing tasks across multi-agent systems requires a modular approach to system architecture, in which development, testing, and troubleshooting are streamlined, reducing disruption. A similar approach to infrastructure can help.

Dell Technologies offers the Dell AI Factory, which brings together AI innovation, services, and a broad ecosystem of partners to help organizations achieve their desired AI outcomes. Dell’s professional services team will help organizations prepare and synthesize their data and help them identify and execute use cases.

Learn more about the Dell AI Factory.