Estimated reading time: 3 minutes

Your only product marketing cannot be “We will replace humanity”.

If you’re an AI company, or something adjacent, “We will do everything your employees can, and a lot more” as a sales pitch will work only so long. It’s late 2026, and that particular track is aging out. The customers will wise up to you. The path forward is a clear product vision, within yourself, and helping customers find specific, revenue-positive use cases.

But why? They already have an omni-product that can seemingly do everything pretty decently. But! An omni-product that can do everything, semi-decently, becomes a commodity when everybody is offering the same thing. AI companies would do well to engage customers in domain-specific product and workflow implementations, and showcase concrete examples of how OpenAI (and co.) would be relevant for them, in explicit, expense-positive way.

For example, if a team wants to implement a classification engine for text analysis, it would be much faster, cheaper, and scalable for them to use existing conventional text classification algorithms, should the usecase be relatively simple, such as customer email routing. OpenAI should train their customers on uding OpenAI offerings to generate synthetic data and try all the alternative classification approaches and implement what might be the best option. It would be a short, and relatively cheap offering for openAI, with their industry-specific forward-deployed engineers, but of great value to the customers, as they’ll be able to understand their needs and solutions much more clearly.

If an insurance company wants to use an OpenAI product to improve the analysis of competitors’ offerings, for example, how might they engage OpenAI offerings? One way, the simplest approach, would be to create a vector-based RAG system and use simple vector querying for simple responses, and then throw in GenAI for improved querying. That is something they could likely figure out by themselves, but what if they couldn’t? Do all of their customers figure it out? Do they need data scientists? What if OpenAI product teams can engage in consulting practices for their customers, telling them these things?

Take a different example. A university might need help with its research labs to leverage OpenAI offerings for better research. OpenAI’s FDE would reach uot to customers to explain how other users use the research function, what the cost/value is like, and how they can extract the most value, and provide common patterns of ai usage. They could assist with the math and jumpstarting the client projects.

The concept of forward-deployed engineers is not new in engineering, or software, so it feels a bit silly for me to explain what feels like such an obvious concept. I suspect it’s going to be a revelation. I know for sure openAI is hiring a few FDE’s right now, but not enough. They should be hiring in the thousands!

OpenAI does not need PR people; it needs product people to engage customers and work with them to save money.