Very quickly, the subject of AI Agents moved from ambiguous concepts to reality. Companies will soon be able to deploy fleets of AI workers to automate and complement – ​​and yes, in some cases supplant – human talent.

“Autonomous agents are one of the hottest and perhaps one of the most high-profile topics in current generation AI,” said Arun Chandrasekaran, vice president and analyst at Gartner, at the press conference. Gartner Symposium/Xpo last week.

However, while autonomous agents are all the rage in the consultancy’s new generative AI hype cycle, he emphasized that “we’re at the very early stage of agents. This is one of the main research goals of AI companies and research laboratories in the long term.

Top Trends from Gartner’s AI Hype Cycle for Generation AI

Based on Gartner’s Hype Cycle 2024 for Generative AI, four key trends are emerging around gen AI, among which autonomous agents are at the forefront. Today’s chatbots are advanced and versatile, but are “very passive systems” that require constant prompting and human intervention, Chandrasekaran noted. Agentic AIon the other hand, will only need high-level instructions that can be divided into a series of execution steps.

“For autonomous agents to thrive, models must evolve significantly,” Chandrasekaran said. They need reasoning, memory and “the ability to remember and contextualize things.”

Another key trend is multimodality, Chandrasekaran said. Many templates started with text and have since expanded to code, images (input and output), and video. A challenge in this area is that “the very fact that they are becoming multimodal, they are also becoming larger,” Chandrasekaran said.

Open source AI is also on the rise. Chandrasekaran pointed out that the market has so far been dominated by closed-source models, but that open source offers customization and deployment flexibility: models can run in the cloud, on-premises, at the edge or on mobile devices.

Finally, cutting-edge AI is coming to the forefront. Much smaller models – between 1B and 10B parameters – will be used for resource-constrained environments. These can run on PCs or mobile devices, providing “acceptable and reasonable accuracy,” Chandrasekaran said.

The models are “getting leaner and expanding from the cloud to other environments,” he said.

On the way to the hollow

At the same time, some business executives say AI hasn’t lived up to the hype. The AI ​​generation is starting to sink into the trough of disillusionment (when technology fails to meet expectations), Chandrasekaran said. But this is “inevitable in the short term”.

There are several fundamental reasons for this, he explained. First, venture capital firms have funded “a huge number of startups” – but they still vastly underestimated the amount of money startups need to succeed. Additionally, many startups have “very fragile competitive moats,” which essentially serve as a cover for a model that doesn’t offer much differentiation.

Additionally, “the fight for talent is real” – think acquisition models – and companies underestimate the scale of change management. Buyers are also increasingly raising questions about business value (and how to track it).

There are also concerns about hallucinations and explainability, and much more needs to be done to make models more reliable and predictable. “We are not living in a technology bubble today,” Chandrasekaran said. “Technologies are progressing sufficiently. But they’re not moving fast enough to meet the high expectations of today’s business leaders.

Unsurprisingly, the cost of building and use AI is another significant obstacle. In a survey by Gartner, more than 90% of CIOS said that cost management limits their ability to leverage AI. For example, data preparation and inference costs are often grossly underestimated, says Hung LeHong, a distinguished vice president analyst at Gartner.

Additionally, software companies are increasing prices by up to 30% because AI is increasingly integrated into their product pipelines. “It’s not just the cost of AI, but also the cost of the applications they’re already running in their business,” LeHong said.

Basic AI use cases

Yet business leaders understand how AI will evolve. Three-quarters of CEOs surveyed by Gartner say AI is the technology that will have the most impact on their industry, a significant increase from 21% in 2023 alone, LeHong noted.

That percentage “is growing more and more every year,” he said.

Right now, the focus is on internal customer service functions where humans are “always in charge,” Chandrasekaran emphasized. “We’re not seeing a lot of customer-facing use cases with Gen AI yet.”

LeHong pointed out that a significant number of enterprise AI initiatives are focused on increasing the number of employees to increase productivity. “They want to use generation AI at the individual employee level.”

Chandrasekaran highlighted three business functions that stand out in terms of adoption: IT, security and marketing. In computing, some uses of AI include code generation, analysis, and documentation. When it comes to security, technology can be used to increase SOC in areas such as forecasting, incident and threat management, and root cause analysis.

In marketing, AI can be used to provide sentiment analysis based on social media posts and to create more personalized content. “I think marketing and AI generation are made for each other,” Chandrasekaran said. “These models are quite creative.”

He highlighted some common use cases in these business functions: content creation and augmentation; summary of data and information; automation of processes and workflows; forecasting and scenario planning; customer support; and software coding and co-pilots.

Additionally, businesses want to be able to query and retrieve their own data sources. “Enterprise search is an area where AI is going to have a significant impact,” Chandrasekaran said. “Everyone wants their own ChatGPT.”

AI is evolving rapidly

Additionally, Gartner predicts that:

• By 2025, 30% of companies will have implemented an AI testing and improvement strategy, compared to 5% in 2021.
• By 2026, more than 100 million humans will interact with virtual robot or synthetic colleagues and nearly 80% of invitations will be semi-automated. “Models are going to get better and better at analyzing context,” Chandrasekaran said.
• By 2027, more than 50% of companies will have implemented a responsible AI governance program, and the number of companies using open source AI will increase tenfold.

With AI now “coming from everywhere,” companies are also looking to assign responsibility for it to specific leaders, LeHong says: Currently, 60% of CIOs are responsible for leading AI strategies. Whereas before the AI ​​generation, data scientists were “the masters of this field,” LeHong said.

Ultimately, “most of our customers keep throwing things to see if they stick to the wall,” he said. “Now they know which wall to throw it at. Before they had four walls and maybe a ceiling to broadcast them on, now they have a marketing wall, an IT wall, a security wall.