The Pivot
In my early working years, I remember being introduced to a company called TIBCO. This company sells enterprise products to help businesses address a variety of use cases. Let us say your company offered loans to customers: the loan process can be very tedious and time-consuming. One of the hurdles in this process was human intervention. Humans need to make decisions, meaning the process depended on human availability or the speed at which they could close a loan request.
This causes a scaling issue because of the limitations of human beings. A method to solve it exists where you can automate the process and limit manual steps or human intervention. TIBCO offered a product that implemented this automated process. They offered companies this product plus a load of other supplementary products to help solve this use case. Back in the early 2000s, TIBCO’s enterprise suite included BusinessWorks, Rendezvous, Enterprise Messaging Service (EMS), Hawk, BusinessEvents, BPM (Business Process Management), BMA (Business Metrics & Analytics), General Interface (GI), and TIBCO Decision Manager, among others. To the creators of these products, they saw a need for an enterprise to use all of them bundled into their enterprise solution. They would convince companies on how they could integrate all their products by producing a glossy demo that showcased all the abilities their solution could solve. Companies would be very interested in this and end up buying a license for all these products since they saw it as a useful list to have.
But after years of experience with TIBCO, companies realised they actually needed only about two products. All the others were either unnecessary or already offered by another company that did the job better. TIBCO believed all their products would be useful, but users found that only two really were.
I bet when all these products were added, TIBCO’s team thought they had created one of the best things in the world. They did not imagine companies would not need most of what they offered. In reality, they created products they themselves would have needed, assuming that if they saw the usefulness, then the business world would too.
We are seeing the same thing happen with AI products today. OpenAI recently released a major version update with the aim of making the models more “intelligent.” They added a feature that allows the AI to decide which engine to use based on the difficulty of a prompt. They saw this as a game changer. They also bundled other AI products like DALL·E for image processing. The update introduced a wide set of new features, including a real-time model router, speed modes (Auto, Fast, and Thinking), personalities, improved coding and design tools through Canvas, an upgraded voice mode, customisable UI, Agent Mode, the Responses API, the Agent SDK, Deep Research, the o-series models (o3, o4-mini, o3-pro), preference fine-tuning, and GPT-OSS.
The backlash, however, is that people feel OpenAI has stripped away the friendly nature of the model. Many users treated ChatGPT as a friend or confidant. They used it to make life decisions and, in some cases, for emotional support. Just like TIBCO, OpenAI’s new model seems designed more for its engineers than its users. They asked themselves how they would use it, rather than focusing on how the public uses it.
The people who use AI as a friend are often laughed at by the general public. But I see why they do it. In a world that silences people for being honest about their views, dislikes, and pleasures, ChatGPT offers a space to be completely honest and vulnerable without judgment. People feel seen, heard, and helped—just like a friend would do.
Maybe OpenAI is focusing on the wrong use case. Maybe the majority of users are already showing them what makes the product great. And maybe, instead of mocking people who seek companionship from a bot, the public should stop and think about why people prefer a chat AI to feel fully human. Maybe it’s not the users who are wrong. Maybe it’s the world we have built that is forcing people to seek help from a bot instead of a friend.


