The AI sector is progressing at an extraordinary rate, with groundbreaking innovations continually transforming the field. A prominent topic among business leaders is AI agents—intelligent systems ready to expedite digital transformation in enterprises.
At South by Southwest, a panel discussion titled “AI Agents and the Future of Human Collaboration” explored how artificial intelligence will alter workplace interactions. The panelists, Nickle LaMoreaux, the chief human resources officer at IBM, and Hannah Elsakr, founder of Firefly for Enterprise at Adobe, agreed that organizations can significantly benefit from the implementation of AI agents.
What distinguishes AI agents from AI assistants?
AI agents build upon the functionalities of AI chatbots or assistants, advancing it further by independently performing actions through reasoning and inference rather than through detailed, step-by-step instructions. LaMoreaux exemplified this by comparing the use of an AI assistant to that of an agent when making a restaurant reservation.
In this scenario, if you request an AI assistant to arrange dinner at a restaurant, it may successfully secure the booking and even extend an invitation to others involved. However, it lacks the capacity to utilize further context to improvise.
LaMoreaux clarified that an AI agent would likely access your calendar, realize your panel ends at 7 p.m., recognize it takes 20 minutes to reach the restaurant, and adjust the reservation to 7:20 p.m. It might also identify that your first choice is booked and instead secure a reservation at your second choice—showing its capability to perform actions autonomously.
“This notion of executing tasks in programmatic manners that can encompass various scenarios, I believe, illustrates the distinction between assistants and agents,” LaMoreaux elaborated.
What advantages do AI agents offer to your organization?
The most apparent advantage of integrating an AI agent into a business is its ability to relieve employees of tedious and administrative tasks, such as HR and backend operations. This enables workers to focus their reclaimed time on tasks that require human creativity and decision-making skills.
“For every minute spent on an HR procedure—whether it be transferring an employee, requesting time off, checking your 401(k) balance, contemplating necessary classes, or evaluating salary increases—it’s a minute that an individual is not engaging with clients, building a product, or being inventive,” LaMoreaux expressed.
The time-saving benefits are clear—after all, time equates to money—but another often neglected advantage is the orchestration aspect of AI agents. This orchestration function relates to how AI agents manage various underlying tools to accomplish tasks and achieve results efficiently.
For instance, if you wished to arrange a trip, an agent could analyze different platforms, such as your calendar to determine available times, the airline application for flight booking, a hotel app for lodging arrangements, and your email to notify your boss about your absence. The ability to integrate all these elements is a vital component of the value AI agents provide to businesses.
Given the plethora of tools available, LaMoreaux noted that employees often struggle to identify which tools to utilize, even when they appreciate the concept. Additionally, last year at IBM, half of employees and managers did not access the HR system, despite it housing nearly a hundred HR tools due to this confusion.
“Don’t overlook the significance of this orchestration factor; many employees and organizations face AI tool overload,” LaMoreaux remarked. “What this agent accomplishes—this orchestration layer—affords employees and managers the ability to access a single interface and interact with all those underlying tools.”
What concerns exist regarding AI agents?
Since the surge in popularity of generative AI, a primary worry has been its potential to affect employment and displace the human workforce. This anxiety is intensified with the introduction of AI agents, as they can perform tasks for people with minimal human input or oversight, further reducing the necessity for humans in specific roles.
However, just because an AI agent is capable of performing certain tasks does not imply it can handle all of them. As previously noted, there are essential business roles that an AI agent cannot supplant, such as engaging in face-to-face client interactions or handling complex tasks that involve problem-solving, experiential knowledge, or creativity.
Indeed, LaMoreaux notes that AI agents may increase the demand for human involvement in certain areas.
“With the advent of AI and generative AI, expertise in specific domains is becoming more crucial, not less,” stated LaMoreaux. “AI agents already possess foundational knowledge, so when unique challenges arise, those will be funneled to humans, who will need to tackle higher-order questions and problems.”
Expanding on this concept, Elsakr emphasized that the models themselves lack original ideas. Therefore, it is essential for humans to generate innovative ideas and dedicate genuine effort to implementing significant concepts.
A further concern about AI agents is their complete autonomy, which allows them to manage vital business tasks and sensitive data without ongoing human oversight, potentially leading to issues with bias, privacy, and mistakes.
Nonetheless, it is vital to remember that AI agents can be customized to meet the specific requirements of a company. They are granted access solely to the information that the organization chooses to disclose and can execute only the actions that they are specifically authorized to perform. Ultimately, these responsibilities should be carefully selected to ensure they align with the company’s objectives and principles.
For instance, LaMoreaux mentioned that IBM utilizes an agent to assist in matching candidates with suitable job openings, rather than employing an AI agent for selection purposes that filter out applications. Although other organizations deploy AI agents for screening and selection, this approach risks introducing AI bias and potentially overlooking candidates from diverse educational backgrounds, which could be at odds with the company’s culture.
This ensures that the company retains control over the behavior of AI agents. Consequently, even though the technology operates autonomously, users can feel reassured that it will not act outside of the permissions granted to it.
What steps should businesses take regarding AI agents now?
While some leaders in business may be cautious about adopting the technology and prefer to observe how AI agents function for others, both panelists concurred that taking action is essential at this time.
“You cannot delay. The only employees likely to be replaced are those not utilizing AI; similarly, the only companies at risk are those not leveraging AI,” LaMoreaux remarked.
At the personal level, even if business leaders are slow to adopt the technology, individuals should familiarize themselves with it and seize the opportunity to learn how to effectively implement and understand it.
Unlike the internet or search engines, which had a more gradual integration into business operations, AI has surged forward at a much faster rate, creating a “tectonic shift,” as noted by Ekstar. Thus, the optimal time to begin the upskilling process is now.
“You don’t have an option here; what I advise people to do, ensuring your CTO approves the tool, is that you must embrace this right now and enhance your skills,” Ekstar urged.
As AI agents rapidly emerge as a foundational force behind enterprise microservices, they will require acquisition, onboarding, and guidance—much like their human counterparts.
The role of agentic AI within interconnected enterprises extends beyond simple helper applications. AI agents are quickly becoming a crucial element behind the microservices that constitute the framework of enterprise systems. Furthermore, as these agents become more widespread, IT departments will resemble virtual “human resources” units, acquiring, onboarding, and managing AI-powered assistants alongside HR’s responsibilities in managing human capital.
These insights emerged from a panel organized by Deloitte at the recent Mobile World Congress, which examined the developing role of AI agents in enterprises. Panelists indicated that agentic architecture mirrors the emergence of microservices architectures, partitioning extensive applications into flexible, independent units.
“Agentic AI represents the next advancement in deconstructing and resolving issues,” expressed Bryan Thompson, Vice President for GreenLake product management at HPE. With agentic AI specifically, there are chances to “utilize these types of models and decompose them into almost a microservice approach to address them—fragmenting them into specialized services.”
Agentic AI facilitates the integration of enterprise workflows, as agreed by Fred Devoir, global head of solution architecture for telecommunications at Nvidia. “We assemble components into a RESTful architecture. Nvidia has optimized these with our microservices, then combined those microservices into templates to provide rapid results or a quick time to value.”
Agentic AI undoubtedly offers capabilities that far surpass those of conventional microservice architectures. “Until now, we’ve never had a technology that could ideate or operate autonomously,” remarked Abdi Goodarzi, who leads generative AI products, innovations, and new businesses at Deloitte. “Consider that remark, along with any software solution you’ve previously encountered. None of them could function independently in any capacity. That is where the true strength of AI lies.”
The services of agentic AI undertake many burdensome tasks typically handled by humans—essentially, a complementary workforce is created but managed by the IT department rather than HR. “Human capital management and agentic AI capital management are essentially the same concept,” Devoir stated. “However, the distinction lies in the fact that instead of having HR for humans, you now have an IT department acting as HR for all these agents.” The IT department also assumes responsibility for “curating, guardrailing, training, and refining AI agents to perform specific tasks and engage with human workflows. This is no trivial undertaking. A significant amount of effort is required for this. It parallels HR, but at a much deeper technical layer.”
This shift also signifies profound transformations within organizations. “Humans possess emotions, whereas agents do not,” Goodarzi explained. “How do you integrate the emotions that will influence the execution of the work? When work is executed differently, cultural shifts must occur, talent strategies modified, and the collaboration between humans and machines must be altered.”
However, transitioning to an enterprise powered by agentic AI comes with its own set of hurdles—particularly concerning data, reliability, and talent. “Organizations have invested heavily in organizing their structured data,” Goodarzi noted. “They’ve built ERP systems and established systems of record and action.” Consequently, these diverse applications lead to isolated data silos.
Agentic AI may help alleviate this issue by allowing deployment of the agent at the data’s location. “Instead of having to aggregate all your data for the AI, you’re deploying the AI to the data,” said Devoir. “When a service call is made, it queries all those data agents for a response and compiles that information into a model.”
Next, there’s the concern regarding the reliability of agents. “One must consider whether the data in question is accurate,” Goodarzi highlighted. “Am I receiving the correct outcomes? Previous technologies were designed to facilitate transactional activities. Agentic AI, however, operates on probabilistic principles. It provides the most probable answer because trained agents possess extensive knowledge on how to process data and render decisions and recommendations.”
The question of trust then arises: “Can I rely on this agent? Is this data trustworthy? Am I facing the correct data? That issue also needs to be addressed.”
In summary, “these are novel concepts for businesses,” Goodarzi stressed. “This is a significant factor in the slow pace of adoption. However, the capabilities are genuine. The technology is sufficiently advanced for use within enterprise production systems. I genuinely believe that this is the year it will gain momentum.”
Leave a Reply