Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
OpenAI is betting that artificial intelligence-powered assistants will “hit the mainstream” by next year as tech groups, including Google and Apple, race to bring so-called AI agents to consumers.
AI agents, which can reason and complete complex tasks for people, have become the newest front in the battle between tech companies as they look to drive revenues from the fast-developing technology.
“We want to make it possible to interact with AI in all of the ways that you interact with another human being,” said Kevin Weil, chief product officer at OpenAI.
“These more agentic systems are going to become possible, and it is why I think 2025 is gonna be the year that agentic systems finally hit the mainstream,” he added.
At its developer day in San Francisco on Tuesday, OpenAI revealed increased access to its new model series called o1, which has improved reasoning, as well as GPT-4o’s advanced voice capabilities. Developers will be able to access this technology in real time, where the AI can understand voice commands and converse in speech in a live scenario akin to a call.
The push to bring AI agents to the masses is one way OpenAI expects its technical advances will help drive future profits, as it moves ahead with plans to restructure as a for-profit company.
The fast-growing start up is aiming to complete a $6.5bn funding round this week at a $150bn valuation by persuading backers it has the capacity to beat its rivals to critical technological milestones and dominate the sector. Investors in talks with the company in recent weeks have included Microsoft, Nvidia, SoftBank and venture capital firms Thrive Capital and Tiger Global, according to people familiar with the discussions.
Microsoft, Salesforce and Workday last month put agents at the centre of their AI plans, while Google and Meta have also indicated this would be a focus for them when putting their AI models into their products.
While AI-powered assistants have been in train for nearly a decade, these latest advances allow for smoother and more natural voice interactions and superior levels of understanding thanks to the large language models (LLMs) that power new AI models.
Last year, OpenAI released “assistants application programming interface”, which was designed to enable developers to build agents using its technology. But the company said this was hampered due to limitations in the capabilities of earlier models.
Weil said OpenAI’s latest models’ improved ability to think and reason would manifest in its products, including ChatGPT, and for start-ups and developers who build products using its API. The company would not comment on whether it immediately plans to build its own AI agent.
One example shown in a bespoke demonstration of the tools was speaking to an AI system to help source products to buy locally, such as strawberries. The AI would then call the business to place an order of strawberries, taking on the user’s instructions for how many and the desired spend.
OpenAI said any uses of such a technology would not be allowed to conceal it was AI rather than a human and was only available to developers in six presets, rather than building new voices.
“If we do it right, it takes us to a world where we actually get to spend more time on the things that matter and a little less time staring at our phones,” said Weil.