The advent of the AI agent

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

As soon as the iPhone was unveiled in 2007, its transformative potential was clear. But it took the launch a year later of the App Store, a marketplace for applications built by third-party developers, to unleash the smartphone’s full power to change how people work, interact, shop and amuse themselves. Evangelists of AI are hoping something similar can happen with generative AI, the technology behind the ChatGPT chatbot. Its creator OpenAI this week announced plans for a “GPT Store”, enabling users to develop and market customised bots tailored for specific functions. Some rivals are doing similar things — marking a new stage in the march of Big Tech: the emergence of AI-powered “agents” that can carry out tasks on behalf of consumers.

OpenAI is releasing a platform for subscribers to its ChatGPT Plus service for developing bots powered by GPT-4, its underlying AI model. Developers will be able to create “GPTs” for external use, as the company is calling these AI apps, for purposes that might range from maths tutoring to interior design to creating presentations. They can be branded and marketed via the coming GPT Store, with OpenAI eventually splitting revenues with the most popular creators.

More notably, GPTs will be able to “plug in” to other websites and services — enabling them to perform tasks such as sending emails or making payments or bookings. The ability to handle online tasks turns AI-driven apps into agents that can, over time, plan and perform more complex actions. Think of a shopping assistant that can browse online for specific products and then buy them, or a concierge that plans and books travel — and which builds up an intimate knowledge of a user’s preferences.

As when smartphone apps came along, it is unclear how big the agents market may become, what it will look like, or even if it will really take off. There is certainly no guarantee OpenAI will dominate the space. Creating its store accelerates the transformation of what was a research lab into a platform business — with at least the potential to be a disrupter of previous disrupters.

Yet while ChatGPT has garnered attention, it has also spurred Big Tech rivals such as Google and Meta to accelerate investment in AI tools — and they have much deeper pockets, even if OpenAI has a wealthy backer in Microsoft. Meta has released a commercial version of its open-source AI model, allowing businesses and start-ups to build custom software on top of the technology, which some observers see as having greater potential.

The advent of AI agents also raises many questions — and, just days after the UK’s global AI summit and US president Joe Biden’s executive order on the technology, highlights how fast regulators may have to run to keep up. Once autonomous AIs are roaming the internet performing tasks, who will police them and how? The rise of well-informed assistants has implications for “surveillance capitalism”. Existing businesses, which may fear AI-driven bots freeriding on their products, services or content, may not give them access — thwarting the rise of agents in the first place.

Whatever happens, OpenAI’s chief executive Sam Altman is clear that being a platform company is not OpenAI’s destination, only a way station en route to his real goal: creating artificial general intelligence, capable of solving profound issues from cancer to climate change. Is the vision he shares with other AI proselytisers of a superintelligence harnessed in the service of a lesser (human) one feasible — or will the reality be more dystopian? As AI begins to carry out tasks autonomously, we will move closer to finding out.