OpenAI's Enterprise Push Looks a Lot Like What Siemens Did in the '90s

What's interesting is the enterprise healthcare angle. Genmab, the biotech company, launched what they're calling "AI Everywhere" with ChatGPT Enterprise. OpenAI's blog makes a big deal about their "commitment to security and privacy," which, I'll be honest, remains to be tested at scale in regulated environments. We don't know yet how HIPAA compliance will hold up when clinicians are feeding patient data into these systems daily.

Here's what I'm watching:

• The federal contract value: OpenAI says "essentially no cost" but the GSA partnership has to have terms. What happens in year two? The pricing structure will tell us everything about their long-term strategy.
• Healthcare liability: When a clinician uses ChatGPT for documentation or research and something goes wrong, who's responsible? This is genuinely unclear.
• The Excel integration: This is sneaky smart. Excel is where actual business decisions get made. Not in fancy dashboards, in spreadsheets. Embedding AI there is, in a way, more important than all the chatbot stuff.
• Enterprise security claims: OpenAI keeps emphasizing security and privacy for enterprise customers. I'd want to see independent audits before I'd trust it with anything sensitive.

The GPT-5.2 announcement (their blog post) talks about "agentic workflows," which is the new buzzword for AI that can actually do things rather than just talk about them. When I was at Kuka, we spent years trying to get robots to handle exceptions without human intervention. The promise of agentic AI is similar: systems that can complete multi-step tasks autonomously.

It's too early to say whether GPT-5.2 delivers on that promise in practice. The benchmarks look impressive but benchmarks always do. Real-world performance in messy enterprise environments is a different animal entirely.

I'm not saying OpenAI's strategy is wrong or evil. It's just business. Microsoft did it with Office, Salesforce did it with CRM, and now OpenAI's doing it with AI. The question for everyone else in this space is whether they can build enough differentiation before the lock-in becomes permanent.

For robotics companies specifically, and this is my beat after all, the lesson is clear. If you're building AI-powered automation, you need to think hard about whether you want OpenAI's models at the core of your stack. Today it's free or cheap. Tomorrow, well. I've seen how these stories end.

Everyone's talking about benchmark scores. I think the real story is what this means for robots that need to think.

Sarah Williams · 2 hours ago · 5 min