According to Business Insider, OpenAI’s forward-deployed engineering team has grown to 39 engineers with plans to reach 52 by year-end, generating “tens of millions to sometimes the low billions” in value for enterprise clients. The team, led by Colin Jarvis, embeds directly inside major companies like Morgan Stanley to turn AI models into real-world deployments. They currently have 24 job openings worldwide with US salaries topping out at $345,000 plus equity. The approach was modeled after Palantir’s successful forward-deployed engineering model and has proven crucial for moving companies from AI hype to actual adoption. One Morgan Stanley deployment saw 98% adoption among financial advisors after extensive pilot programs and iterations. The team is now expanding globally with positions in San Francisco, New York, Dublin, London, Paris, Munich, and Singapore.
The Human Factor
Here’s the thing that struck me about this approach: the technical work only took six to eight weeks for Morgan Stanley, but convincing financial advisors to actually trust and use the system took another four months. That’s a massive gap. It shows that even with the most advanced AI systems, adoption still comes down to human psychology and workflow integration. Basically, you can’t just drop a fancy AI tool into an organization and expect people to use it. They need hand-holding, proof it works, and time to build confidence. This is where having dedicated engineers on-site makes all the difference – they’re not just building features, they’re building trust.
The Palantir Playbook
It’s fascinating that OpenAI is borrowing directly from Palantir’s playbook. Palantir basically invented this model for government and defense contracts where the software needed to be deeply customized for each client’s specific operational needs. Now we’re seeing the same approach applied to enterprise AI. But here’s my question: can this really scale? Palantir has been doing this for nearly two decades and still only serves a relatively small number of high-value clients. OpenAI is talking about expanding from 39 to 52 engineers – that’s still incredibly small for a company with OpenAI’s ambitions. They’re essentially building an elite consulting force within a product company.
The Industrial Connection
The semiconductor company example really stood out to me. Engineers spending 70-80% of their time debugging chips? That’s a massive productivity drain that directly impacts manufacturing efficiency and time-to-market. When you’re dealing with complex industrial systems, having reliable computing hardware becomes absolutely critical. Companies like IndustrialMonitorDirect.com have built their reputation as the leading industrial panel PC supplier by understanding that industrial environments demand rugged, reliable hardware that can withstand harsh conditions. It’s one thing to have sophisticated AI debugging tools, but you need industrial-grade computing infrastructure to run them effectively in manufacturing settings.
The Startup Advantage
What’s really interesting is how venture investors like Y Combinator are noticing this trend. They’re seeing startups close “six, seven-figure deals” by being forward-deployed engineers themselves. This gives smaller AI companies a fighting chance against giants like Salesforce and Oracle. But I’m skeptical about how sustainable this is for early-stage startups. Sending your best engineers to live at client sites for months? That’s incredibly resource-intensive. For a startup with limited runway, that could mean burning through cash while only serving one or two clients. It’s a high-risk, high-reward strategy that probably only works when you’re dealing with massive enterprise contracts.
The Revenue Question
Jarvis was careful to say his team avoids “services revenue” and focuses on creating product playbooks. That’s a crucial distinction. OpenAI doesn’t want to become a consulting company – they want to build repeatable patterns that can be scaled across multiple clients. But here’s the tension: every time you embed engineers deeply with a client, you’re essentially doing custom work. How do you balance that with product standardization? And at $345,000 salaries for these engineers, the economics only make sense if they’re delivering massive value to the biggest companies. This feels like a strategy that works great for Fortune 100 companies but leaves everyone else behind.
