Allow OpenAI to reintroduce itself.
After watching an upstart (Instinct) and tech giant's (Meta) AI agents grab the spotlight, OpenAI has responded. Dots, OpenAI's "always on" AI agent, got the big reveal at the startup's annual DevDay conference in San Francisco.
It's the third major AI agent launch over the past few months and could be tech companies' best bet at getting the customer lock-in they've been chasing.
The first noticeable difference between Dots and Meta's Muse and Instinct is that it's not free. Initially, Dots will be available to Enterprise and Business Premium users and ChatGPT Pro subscribers.
It wasn't all about AI agents, though. CEO Sam Altman outlined his three-part plan for OpenAI focused on models, building, and distribution. OpenAI also made several other announcements, including ChatGPT Space, a new way to collaborate with coworkers, and a new top-tier plan that'll run $500 per month.
BI's Stephen Council was at the event and wrote about the big takeaways. Now that we've got more agents than we know what to do with, what are we supposed to use these things for?
I haven't jumped into the AI agent fray … yet. It's less about data security and more about resisting the complete handover of my life to the bots. Whether it's AI atrophy or getting addicted to a service that could suddenly start costing me money, I see plenty of risks.
Plenty of executives aren't hesitating. BI's Hugh Langley and Polly Thompson spoke to bosses at Anduril, KPMG, and Adobe, among others, about the tasks they are delegating to AI agents.
Lots of the most useful examples circulating are about canceling forgotten subscriptions or negotiating a better price. But once those pesky subscriptions disappear — hopefully not this newsletter — what will keep people coming back?
BI's Peter Kafka recently raised this issue. There's lots of speculation AI agents could reshape the commerce business, taking a fee on those transactions. But a MoffettNathanson report suggests the average person only has 11 monthly payments that an agent would likely handle.
Personal agents could finally be the hook that keeps users committed to one AI company. I've talked about an era of chatbot monogamy. Unfortunately for AI companies, it's too easy to diversify tasks across different chatbots.
Agents are a different story. It's one thing to drop a question or a request into multiple apps. It's another to give several AI agents access to a bunch of your data and let them run wild. (The hardcore techies might disagree, but understand you're the exception, not the rule.)