As early as 2025, the industry will begin to shift from offering AI assistants to providing full-fledged AI agents capable of completing entire tasks on their own.

The tech industry is poised to begin offering GenAI-based capabilities in 2025 that are more akin to a full-fledged agent than the types of assistants we see today, according to the CEO of a large digital engineering company focused on AI projects.

Rather than AI turning us into “cyborgs,” “the more likely future is that each of us will have a Jarvis-like assistant who will likely give us superhuman abilities,” said Asif Hasan, co-founder and president of Quantiphi, referring to the famous “Iron Man” AI system. “If we do it right, that’s the kind of future we can look forward to.”

(Related: Quantiphi CEO on Four Popular GenAI Use Cases and Google’s AI Push)

Hasan spoke Monday at the XChange Best of Breed 2024 conference, taking place in Atlanta and hosted by CRN’s parent company, The Channel Company. He noted that Quantiphi, based in Marlborough, Massachusetts, has completed 2,500 AI projects to date and that almost all of its revenue at this point is related to AI applications or AI readiness. ‘AI.

The next phase of GenAI’s advancement is expected to begin as early as next year, Hasan said. “I believe 2025 will see what we call the rise of agents.”

While today many AI systems are already capable of synthesizing information, the next step is for these systems to start completing entire tasks for a user, such as booking an airline, a- he declared.

This will require AI systems to acquire better reasoning skills, including the ability to plan a task and break it into smaller pieces. AI agents will also need the ability to browse the web, so this type of integration will be crucial, according to Hasan.

Ultimately, however, “we think 2025 will be the year we start to make significant progress in this direction,” he said.

Ahmed Mahmood, CEO of Ocean Solutions, a solutions provider based in Ashburn, Virginia, said there is no doubt that the future will rely on AI systems that can handle more complex tasks. In addition to the AI ​​systems themselves, data systems will likely need to be “fine-tuned” so that it is possible for systems to interact more effectively with user commands, Mahmood said.

It might also be possible to bring together multiple AI systems to handle different parts of a task, he added.

“When you have a complex need, AI can always help,” Mahmood said.

For Quantiphi itself, steps in this direction include developing a software engineering agent that can take project tasks and break them down into “very atomic units” before presenting them to developers, said Hassan.

As a result, “one person can complete an entire task in a week, without having to talk to anyone,” he said, noting that this means developers have no downtime while waiting to be integrated into the next one. project. “The task is self-explanatory. They are able to complete the task, check the code, and use that time productively.