How to Recognize ‘Agent Washing’ Before AI Leaves You out to Dry

13 hours ago 4

Companies of all sizes are being deluged by vendors who want them to deploy the artificial intelligence (AI) trend du jour: AI agents.

But most of the AI agent systems being sold today are not truly agentic, according to a report from business research and insights firm Gartner. According to the report, out of thousands of AI agent systems touted by vendors, only 130 are real.

Agentic AI, or systems with the autonomy to plan, reason and act toward goals with limited or no human input, is being conflated with simpler tools that lack those capabilities.

“Many vendors are contributing to the hype by engaging in ‘agent washing’ — the rebranding of existing products, such as AI assistants, robotic process automation (RPA) and chatbots, without substantial agentic capabilities,” Gartner wrote.

Vendor selection is critically important. When CFOs were considering deploying generative AI, 25% said reliance on vendors was one of the drawbacks to integration, according to a PYMNTS Intelligence May 2024 report. Fast forward a year: CFOs have become much more comfortable with generative AI but are unsure whether agentic AI is “battle-ready,” according to a July PYMNTS Intelligence report.

Gartner predicted that more than 40% of agentic AI projects will be canceled by 2027 due to high costs, unclear business value and weak risk controls brought on by AI systems incorrectly marketed as agentic.

“True AI agents are defined by goal-driven autonomy — the ability to work dynamically and proactively, with self-determination, to pursue long-term business goals,” Sagi Eliyahu, co-founder and CEO of the tech orchestration platform Tonkean, told PYMNTS.

These agents can use tools, taps unique skills as well as other agents to complete complex work across tech environments, Eliyahu said.

But the key ingredient is integration, Eliyahu said.

“A true agentic AI system orchestrates agents across every relevant piece of technology or team environment. If the ‘agent’ only handles discrete tasks that are defined by the user, if it only works inside its own system, or if it’s only accessible through chat, it may in fact be an AI capability, but it’s not an agent — it’s an automation or it’s a chatbot.”

Akhil Sahai, chief product officer at Kanverse.ai, wrote in a post on LinkedIn that companies need to ask the following questions to identify whether an AI system is truly agentic:

Can the system operate without constant human input?
Does it pursue goals autonomously rather than follow scripted tasks?
Can it reason, plan and improve with experience?

“If the answer to any of these is ‘no,’ it’s not an AI agent,” Sahai said.

Demand for Hard Evidence

Eliyahu said it’s important that an agentic AI system supports orchestration. This refers to several bots coordinating on tasks to achieve a common goal, by themselves or with minimal human intervention. In traditional automation, the bot does one task at a time in a fixed sequence. It can’t plan, adapt or collaborate.

“Orchestration is the essential infrastructure for leveraging AI agents to their full potential,” Eliyahu said. “Agentic orchestration is how you instrument AI agents for enterprise. It’s an approach that puts agents alongside employees to coordinate workflows, execute tasks and drive outcomes — all while following configurable policies and guardrails.”

But the problem is that the term “AI agent” today has become part of a trend. “Everywhere you look, companies are branding their products as ‘AI agents,’” Sahai added. “The term has become a catch-all, slapped onto everything from basic workflow automation to applications that simply call an LLM for a response … This isn’t harmless marketing; it’s fueling confusion about what AI agents truly are.”

Sahai compared the trend to earlier rounds of “cloud washing” and “AI washing,” in which vendors relabeled legacy software to appear modern. He noted that simple rule-based automation, apps with embedded language models and tools that require frequent human intervention are often mislabeled as agentic systems.

These exaggerated claims come at a cost: erosion of trust, confusion, stagnation and wasted investment.

“Most agentic AI propositions lack significant value or return on investment (ROI),” said Anushee Verma, senior director analyst at Gartner, in the report.

Despite current shortcomings, Gartner sees long-term potential for agentic AI. It predicted that by 2028, 15% of daily work decisions will be executed autonomously, and 33% of enterprise software applications will include agentic capabilities — up from less than 1% today.

But caveat emptor must apply.

“Don’t settle for AI agent as a label,” Sahai said. “Demand evidence. Ask hard questions. And avoid repeating the cycle we saw with cloud-washing and AI-washing.”

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

Meet the AI Agents Acing Compliance Tests and Correcting Government Data

AI Agents Do Well in Simulations, Falter in Real-World Shopkeeping Test

Agent washing is widespread, with vendors mislabeling basic automation or LLM apps as AI agents, creating confusion and leading to wasted investments.
True AI agents can work together to achieve a shared goal with minimal or no human intervention. They can reason, plan and adapt.
To avoid falling for inflated claims, business leaders should demand proof and ask critical questions.

Read Entire Article