Who is the real change agent in the race for agentic artificial intelligence (AI)?
In less than a year, dozens of companies — ranging from the usual suspects like Microsoft Corp. and ServiceNow Inc. to unlikely sources like PwC and Zoom Communications Inc. — have announced AI agents to do just about anything. Some agents will help workers find potential business opportunities, then set up and book meetings with prospective clients, according to Daniel Newman, CEO of The Futurum Group. Others, like voice AI agent Ribbon AI, is conducting interviews and making job recommendations to human recruiters.
The hype over AI agents has reached molten-lava hot status, as evidenced by a Futurum report that agentic AI will tackle up to $4 trillion worth of human labor globally by 2030, paralleling predictions from NVIDIA Corp. CEO Jensen Huang of a “multitrillion-dollar opportunity” with billions of automated digital workers.
Chief information officers “now prioritize AI-driven automation to replace mundane tasks and augment existing teams with digital labor,” Futurum’s report said. “This shift is far more than a technology refresh. It’s a profound evolution of work itself.”
No surprise then that the current massive scrum resembles a free-for-all, with nearly every B2B company chasing a potential gold rush with a definite FOMO (fear of missing out) as motivation.
Indeed, investments in AI solutions and services are projected to yield a global cumulative impact of $22.3 trillion by 2030, or about 3.7% of the global Gross Domestic Product (GDP), according to a recent IDC report.
“While AI copilots have already made inroads across industries, the next evolution — autonomous agents with greater decision-making scope — is arriving quickly. AI agent startups raised $3.8 billion in 2024 (nearly tripling 2023’s total),” according to a CB Insights Research report.
The land rush is constant, with some aspirants more prolific than others: Salesforce Inc. seemingly carts out a feature weekly. AI agent has become so ubiquitous, in fact, that it’s lost all meaning, leading to big promises and even bigger letdowns, warns Esteban Sancho, chief technology officer of North America at Globan.
Granted, a dizzying array of options has left CIOs and CTOs “de-risking and identifying which of these models we will apply to the organization and leverage and how and at the pace that it is going, most companies simply don’t have time to test them all,” Chris Mattmann, chief data and AI officer at UCLA, said in an email.
“We’re seeing an explosion of AI agent announcements, but the reality is that few are truly enterprise-ready,” Sara Hanley, vice president of growth marketing at ASAPP, said in an email. “Most solutions are still demos or early stage co-pilots and not production-grade systems. That definitely creates confusion for IT and CX leaders that are trying to understand hype from signal.”
Things have gotten so overwhelming, Kibo Commerce launched an Agentic Commerce initiative for enterprise retailers coping with fragmented data and decision fatigue. Rather than positioning itself as just another AI agent vendor, Kibo is embedding AI agents into a modular commerce and OMS (Office Management System) stack to perform autonomous, adaptive operations across fulfillment, personalization and decision support.
Finding — and Succeeding — With the Right AI agents
Despite bold boasts from tech giants, most AI agents are destined to fail when initially deployed in real-world enterprise settings, tech consultants warn. Current AI systems struggle with data readiness, security and operational complexity, making them ill-equipped for mission-critical business functions in industries like retail, finance, IT and manufacturing.
Bugcrowd CEO Dave Gerry predicts 80% of AI agent vendors will be gone the next few years, victims of either lack of success or consolidation.
“From talking to customers and prospects over the years, I can say confidently that decision makers really don’t understand AI, assume it’s magic without knowing how it works and where to even start,” Greggory Elias, founder and CEO of Agents for Hire, said in an email. “A lot of businesses are looking to adopt AI solutions but are overwhelmed and until they see results that are significant, or a solution that is easy to onboard. The 6.3 million or so 500-person and -under businesses are going to wait. I think massive adoption is much further away than people realize.”
“No one is making customized agents simple,” Elias said. “When they do, adoption will skyrocket.”
Adding to the confusion is a general disconnect — so far — between makers of AI agents and their enterprise customers.
Jason Zeng, a data integration director, compares what is going on with AI agents to the automation wave of a decade ago. While he praises them as “great,” he says it’s only applicable by someone who both understands AI and the industry the specific agent is built for.
“The main problem is that many startups create an AI agent, but it’s not designed for any specific purpose,” Zeng said. “The buying company, on the other hand, doesn’t know how the agent should be used for their business.”
Many so-called AI agents in the market today are little more than glorified task managers limited to executing basic, pre-programmed actions without the ability to reason, adapt or operate with true autonomy, claims Barry Cooper, president of the CX Division at NICE. That kind of “exaggerated marketing,” he warns, “not only misleads enterprise buyers but risks undermining long-term trust in AI altogether.”
IT decision makers will only be able to create an effective roadmap if they have a strong point of view around how AI should be deployed across teams and use cases, according to Mayank Bawa, founder and CEO of WorkSpan.
To be fair, there will eventually be a number of winners in a wide range of business categories, but today’s products “have yet to deliver on the promise of agentic AI: true workflow automation,” Alex Salazar, co-founder and CEO of Arcade.dev, said in an email. “Most of these agents are wrappers on LLMs (large language models) that do not meaningfully enhance workplace operations.”
This is largely due, he said, to the “auth problem.” Authentication and authorization are the connective tissue that can bridge LLMs and essential business systems which, so far, have remained isolated from one another. “Integrating AI models with the tools we use every day has been a hurdle that has made building and deploying AI agents both time- and cost-intensive,” he said.
Of course, in the end, the market(s) usually come down to the same factors.
“As for a clear early winner, the incumbents have brand recognition but are still layering generative AI on legacy and rules-based systems,” Hanley said. “Newer entrants are built from the ground up for generative AI automation with real deployments. The winners will be the ones that solve for trust, safety, integration, and real business impact.”
