Datadog sold what startups wanted. You want an observability platform, they would sell it to you piece-by-piece. This was great to the higher-ups; in an era when data science and dashboards were all the craze, every startup you went to had a monthly Datadog bill in the thousands. They were hooked.
Then the economy went to shit, and you know what everyone realized? Datadog was useless. They were taking our money hand-over-fist to give us features that didn't matter and detracted from developer funding. When the higher-ups decided to stop trend-chasing, they made the call to drop Datadog and usually replace it with something else (some lumbering Prometheus stack).
The moral of the story isn't that you can see a need and fill a need; the moral is that people will abandon half-measure solutions. OpenAI isn't filling a need, they aren't directly addressing a problem people have been demanding a solution for. They're selling a trend that you can join-in on for a per-call fee from their API. Much like Datadog, one day startups will start removing ChatGPT to replace it with local LLMs, and then probably remove LLMs altogether later down the line. Mark my words, the most successful businesses that come out of the next 10 years will barely even mention AI as a footnote in their product description.