Why We Need Forward Deployed Engineers
There is a gap between what AI can do and what companies actually get out of it, and the forward deployed engineer is the role the industry needs to close it.
In theory the benefits of AI are obvious. In practice almost nobody on either side of the table knows how to actually apply it. The providers know what their models can do. The companies know what their problems are. The bridge between those two is still mostly missing, and a job title has appeared to plug exactly this gap.
Anthropic recently published findings showing that outside coding agents, most AI agents in the market are still not delivering the returns people expected. That matches what we see directly. The industry is a few years old at most. The taste for where AI fits, where it does not, and what shape the solution should take has not had time to develop. There is plenty of advice floating around and a lot of it is bad. The Uber CTO publicly admitting they burned a year of AI budget in three months is not an isolated story.
The forward deployed engineer exists to close that gap from the inside. The important part to understand is that this is mostly not a development role. In my experience the actual coding is maybe ten percent of the work if executed correctly. The rest is sitting close enough to a specific business to understand its problem domain at a level the provider never will, and figuring out which slice of that domain AI can usefully touch.
That is why the role has to be embedded. You cannot do it from the outside with a deck and a discovery call. You have to be in the room while the work is happening, watching where the friction actually lives, what the team already does well, and what kind of AI shape would produce a real return rather than a demo. The engineering shows up at the end, once the problem is understood. Show up with engineering first and you get the Uber outcome.
The deeper point is that this is a problem for the providers as much as for the adopters. Time will smooth it out over years. On a quarterly basis it is a hard block, which is why the big labs are paying real money to put their own people inside customer organizations.
The framing that helps is to stop asking how to add AI and start asking how to leverage AI here, specifically. Not replacement. Augmentation and improvement of work that already exists, applied to a problem that has been named precisely. Coding agents are one of the few areas where the industry has figured this out well enough that the answer looks obvious. Almost every other domain is muddier, and that muddiness is exactly what the forward deployed engineer is paid to break through.
This is also why we built ChatBotKit the way we did. Forward deployment lives or dies on how quickly you can try an idea without committing to building the system behind it. Building too early is the classic failure mode. The idea has to be identified and validated first, and there is usually a long stretch of back and forth before anyone even understands the problem clearly enough to commit code to it. A platform that compresses that experimentation loop is what lets the role work in practice rather than just on a job description.