I came across Andrew Ng's LinkedIn post on the rise of the AI Forward Deployed Engineer (FDE), and it stuck with me long after I read it. I have a lot of respect for Andrew. He was my teacher at deeplearning.ai and remains a role model for me on the technical side of AI. One thing I have always admired about his work is how consistently he champions what I call the "hard side of AI", the engineering, the systems, the agentic frameworks. But reading his post, I noticed something missing: the soft side. Put it another way, Andrew Ng was describing the future of AI jobs entirely from the technical side. Who builds the systems, who tunes the workflows, who codes the agents. What about everyone else? What about the accountant who has never written a line of code but needs to know how to use AI to close the books faster? What about the school administrator, the policy officer, the small business owner, the mid-level manager who will never touch an agentic framework but whose daily output could double if someone simply showed them how to think with AI rather than around it?

That gap is where I think a second role belongs. I call it the AI Forward Deployed Augmentation Specialist, or FDAs.

What an FDAs actually does

Where an FDE is embedded to build and integrate, an FDAs is embedded to teach, translate, and elevate. The FDE asks, "how do I get this AI system to do what this client's infrastructure requires." The FDAs asks, "how do I get this client's people to think, work, and decide better because AI now exists." One is an engineering function. The other is a human capability function.

In practice, an FDAs spends their time on things like:

  1. Mapping a team's actual workflow and finding the three or four points where an AI tool removes real friction, rather than handing people a chatbot and hoping for the best.

  2. Running short, role-specific training so a procurement officer, a customer support lead, and a finance analyst each learn the version of AI fluency that matters to their job, not a generic tutorial.

  3. Building simple habits and prompting patterns that stick, the way a good coach builds a repeatable routine rather than a one-off trick.

  4. Measuring whether people are actually faster, clearer, or less stressed after adoption, and adjusting the approach when the answer is no.

  5. Sitting close enough to the work to notice resistance early, the colleague who quietly avoids the new tool, and addressing it with patience rather than a mandate from above.

None of this requires writing code. It requires understanding both the tool and the human being asked to use it, and being good at closing the distance between the two. In addition, it is tempting to treat productivity and adoption as a secondary concern, something that happens automatically once the technical work is done. I think that gets the order backward in most organizations, and especially in the contexts I work in across public institutions and growing economies. A government agency or a small enterprise can license the most advanced AI system available and see almost no change in output if nobody on staff knows how to use it well. The bottleneck is rarely the model anymore. It is fluency, trust, and habit. Those are human problems, not engineering problems, and they need a dedicated specialist the same way technical integration does.

There is also a quieter argument for the FDAs role. Andrew Ng noted that FDEs need communication and business skills on top of technical depth, which is part of why they are valuable and somewhat rare. The FDAs flips that requirement. The technical depth required is moderate, enough to understand what the tools can and cannot do, but the communication, coaching, and change management skills are the entire job. That is a very different hiring pool, and it opens AI-era careers to people who came up through training, HR, education, or operations rather than computer science. Given how much larger the global workforce of non-engineers is compared to engineers, the addressable job market for FDAs could end up dwarfing both the FDE and AI Engineer categories combined.

Where this leaves us

Andrew Ng is right that the jobpocalypse narrative is wrong, and that AI is creating new categories of work rather than simply erasing old ones. I just think the picture is incomplete without a role dedicated to the human side of adoption. The FDE builds the bridge. The AI Engineer builds the software that crosses it. The FDAs makes sure the people on the other side actually know how to walk across, and walk across well, rather than standing at the edge wondering if it is safe.

I floated this idea, AI Forward Deployed Augmentation Specialist, as a comment on Andrew Ng's original post, and I want to put it down properly here because I believe it names something real that a lot of organizations are quietly missing. If the next decade of AI work needs engineers to build the systems, it will need augmentation specialists in equal measure to make sure the systems actually get used.

What do you think? I would love to hear whether you have seen this role already taking shape inside your own organization, under whatever name it currently goes by.