Forward Deployed Engineers: underscoring human value in the AI era

Andy Nguyen: While more traditional approaches to engineering typically engage the expert only after others have defined the requirements, the FDE works alongside the customer from the start. This is critically important when deciding where and when to deploy AI during the process of modernizing an IT estate to address an organization’s business objectives. It’s the difference between coming in to build a structure after the customer has decided they need an office tower and working with the customer from the ideation stage to determine what kind of structure they need before optimizing its design and construction.

In other words, the customer may require a structure that accommodates a certain number of people. FDEs work with Kyndryl Human Systems Architects (HSAs) to determine whether that structure should be a stadium or a skyscraper. And then we build and manage it. Technically, we provide a “service.” But results are the focus of everything we do.

This may sound a bit circular, but AI enables us to design the integrated and multifaceted systems that use AI to analyze and capitalize on a customer’s data. We use AI to help us help the customer obtain maximum value from AI. During this process, there’s no getting around having a foundation of expertise in technology and business, along with an understanding and appreciation of the human factors in digital modernization.

An FDE then identifies and assigns the appropriate tasks to AI so the FDE can focus on matters requiring human perspective and judgment. Being able to have AI carry the appropriate loads frees people to employ nuance and discernment across every aspect of the project lifecycle.

Nguyen: Digital technologies are maturing faster than ever, but people are analog. We need proof of safety before we bestow our trust. Everyone alive today is a descendant of people who didn’t run into the dark cave before knowing what was in there, so mistrust is part of our firmware.

That’s why a major part of our jobs as AI and IT systems experts is to be empathetic enough to acknowledge that mistrust exists, while being conversant enough to explain to others how we’re incorporating humanistic principles, safety and governance into the DNA of AI. One of the characteristics that distinguishes us is our Kyndryl Way culture. Empathy is a starting point for us, not an add-on. It’s one of the reasons customers come to us and why — just as important — they partner with us for years or even decades.

We’re using AI to help us build prototypes faster than ever. But we never release something into the wild without proving, and then confirming, that it can function safely and responsibly at enterprise scale. We have an advantage in that Kyndryl Bridge provides us with millions of monthly insights from a variety of critical-infrastructure enterprises. These insights give us real-time visibility into what’s working and how to replicate success quickly, safely and cost-effectively for all of our customers.

The guiding principle in all of these endeavors is “honesty over confidence.” We design our systems to acknowledge when they don’t know something instead of guessing and making mistakes. There is always accountability regarding which data we’ve used, how we’ve made decisions and whether we can validate those decisions. FDEs (and HSAs) build and maintain the types of customer relationships that give us trusted access to explain our technologies in a human context. Building that trust helps remove the barriers between concept and implementation, resulting in stronger returns on AI investments for our customers.

Nguyen: Systems thinking is critical. An FDE needs to understand the complexities of how technology, business processes and people interact. You need to be curious, empathetic and flexible because the field changes quickly. I guess this also means you should have polymathic tendencies, with a strong technical foundation (although AI can help with that), an excitement for learning and the ability to work effectively with a broad range of people from different cultures and areas of expertise.

When it comes to building agentic solutions specifically, there are two approaches gaining real traction in the AI space that I think every FDE should adopt. The first is spec-driven development. An agent will happily produce a large amount of plausible-looking work, but it cannot read your mind, so the time you spend writing down the intent, the constraints and the acceptance criteria before anything is generated is the most valuable time you will spend. A big part of the FDE’s job becomes translating a messy business problem into a specification that is precise enough for an agent to act on, and clear enough for a business stakeholder to recognize and sign off.

The second is an eval-first approach: decide what “good” looks like before you build — not after. Agents are non-deterministic and small errors compound across many steps, so a handful of successful tests tells you very little. You want to write the evaluations up front, grade the outcome rather than the exact path the agent took and turn every real-world failure into a new test case. Together, those two habits are what turn a promising prototype into something a customer can actually trust in production.

Ultimately, the role requires the ability to work with both breadth and depth to identify and solve complex problems. For the right person, it’s exciting, energizing and a lot of fun!