According to analyst Gartner's global AI Research Erick Brethenoux global chief of analyst Gartner, “AI is not doing that job today, so we should leave it alone.”
Today at the company's Data & Analysis Summit in Sydney, Australia, Bretenu said that there was no time to read the summary of the meeting two or five years ago before the creation of such documents became a critical application for generating AI.
“We don't have time to perform five actions in the summary,” he added. “I know what I have to do.”
Brethenoux also questioned why AI would list action items for a meeting instead of doing the work itself, instead of doing the AI.
“Just go and do it already,” he said, urging AI to simplify users' lives by automatically performing boring tasks.
He cited a use case at US healthcare company Vizient. The CTO regularly asked employees what tasks they were troubled with. Armed with feedback from thousands of employees, the company automates its most completed chores.
result? “There's zero immediate recruitment, change management issues,” Bretenu said. The employees then acquired AI and began making good proposals for AI-enabled automation.
Analysts labeled this approach as “empathy AI,” citing another example in real estate companies. This requires 17 steps to assess whether or not you want to rent property to a future tenant.
The company evaluates candidates in each of the 17 steps in turn, so the failure of Step 16 meant that it wasted time in the previous 15 steps. ORG uses AI to automate every step in parallel.
Enterprise Tech vendors are currently advocating automation to use AI agents (bots that can perform several tasks independently) to automate the boring parts of IT OPS or act as a kind of personal assistant.
Brethenoux believes that tech buyers must get their vision with a ton of salt for two reasons.
One is that AI agents are not new. He said industrial companies have been using them for decades in relatively closed systems. Currently, we rely on agents for specific tasks, but software rarely can handle extremely complex tasks.
However, vendors suggest that individual AI agents can easily manipulate many data sources across the enterprise, automatically determine that workers need to attend meetings, and place the meeting in Outlook or Google Calendar.
“Now you have 50,000 agents running around the Enterprise,” he assumed. “How do you adjust this? How do they negotiate?”
Brethenoux said it asked vendors how such automated scheduling would consider the competing needs of employees' superiors, partners or children. Their response, he said, was “silence.”
Analysts believe that vendors and users do not give sufficient consideration to how to build an agent system that addresses these issues.
“This is a software engineering issue,” he said. “The people who are breaking down the system, when they can communicate, they need the degree to which they communicate, the different levels of autonomy they provide within the agent.”
Software engineers also need to decide which information agents can perceive, what they can control and what they can perform.
“That's not a trivial thing,” he said.
The vendor knows this, he said, but it still promotes the idea that Agent Nirvana is within reach.
Brethenoux said the current wave of AI hype will be fueled in part by the fusion of the terms “AI agent” and “generated AI” and the use of both fuzzy definitions.
He lamented the practice by sharing a saying stemmed from French philosophers and author Albert Camus. “Marrying things is to contribute to the misery of the world.” ®
