OpenAI just bought a personal finance startup and a media company. Not because they're suddenly getting into financial advisory or journalism, but because they have a more fundamental problem: The world's most powerful AI model doesn't retain users.
That's the real story behind these deals. And it's more relevant than the acquisitions themselves.
#A subscription nobody needs
ChatGPT Plus costs $20 a month. For many users, that's an expense that gets reevaluated with the next paycheck. Canceling takes 30 seconds. The reason to stay? Usually: habit. And habit is a thin foundation for a business model.
That's exactly what TechCrunch describes in their analysis of the OpenAI deals: The Hiro acquisition aims to build a product with "more hooks than just a chatbot," something that embeds itself deeper into users' lives and thus becomes harder to cancel. Personal financial data, spending behavior, savings goals. That's information you don't just casually take to another provider.
This isn't coincidence. This is retention design.
#Retention isn't a feature you bolt on later
Here's the misunderstanding I see over and over, with clients, with startups, and apparently with one of the world's most valuable companies: You build the product first, then think about why people should stay.
That doesn't work. Not with an app. Not with a website. And not with an AI subscription.
User retention doesn't emerge from better features. It emerges from embedding into life. Through habits that the product helps shape. Through data that gets lost when you switch. Through a community that isn't portable. Through emotional significance that builds over time.
OpenAI has none of that. At least not yet. ChatGPT is useful, sometimes indispensable, but it's not part of daily life the way Instagram or WhatsApp are. You open it when you need something. Then you close it again. That's not a hook. That's a tool.
#The TBPN deal is more honest than it looks
The second acquisition is a business talk show format, a "new media" company. At first glance, this seems like a distraction or an expensive toy.
But Sean O'Kane from TechCrunch nails it: OpenAI's public image has been anything but good in recent months. Leadership conflicts, safety debates, the dispute over converting to a for-profit company. Anyone who understands a company's public perception knows: Earned media is slow and uncertain. Owned media is controllable.
This isn't journalism. This is reputation infrastructure.
Both deals together tell the same story: OpenAI is currently trying to solve two things that no language model in the world can solve. Trust and retention.
#What this has to do with your product
I work with SMBs. No billion-dollar valuations, no VC rounds. But the pattern is identical.
A craftsman comes to me, great references, fair prices, real quality. But the website is a digital business card, and after the first job, there's no reason to come back. No newsletter, no maintenance contract, no customer account, no reminder. The customer is gone, and the next one has to be convinced from scratch.
Or a software provider who explains to me that their product is technically better than the competition. Maybe so. But the competition is deeply integrated into customers' workflows. Switching means data migration, retraining, risk. This isn't a quality problem. This is a switching cost problem, and the competition built those costs deliberately.
The best product doesn't automatically win. The product that's hardest to leave wins.
#What you can actually do
If you have a product or service that stands on a single use case, this question is worth asking: What happens when a customer stops?
If the answer is "nothing special, they just stop," then you don't have a retention problem. You have no retention design.
A few approaches that work and don't require million-dollar acquisitions:
Data that gets lost when switching. If your product learns something about the user and that knowledge isn't portable, you automatically create a switching barrier. This can be preferences, history, personalized recommendations.
Habits that your product helps shape. Daily check-ins, weekly reports, reminders. Not as spam, but as real value that recurs rhythmically.
Embedding in existing workflows. The deeper your product integrates into daily work, the higher the switching costs. Integrations, API connections, exportable formats that are still proprietary enough.
Emotional significance. This sounds soft, but it isn't. When your product is connected to a success, a milestone, a memory, it's no longer just a tool.
#The real question
OpenAI is currently buying answers to questions they should have asked earlier. Not: How do we build the best model? But: Why should anyone stay?
This isn't a technical problem. Anthropic has good models too. Google too. Meta too. The race for the best model is one OpenAI can't win alone, at least not permanently.
User retention is a design problem. And design isn't post-work, it's the question that must be asked at the beginning.
Those who understand this now don't need to make startup acquisitions to solve it.
