The Forward-Deployed Engineer's Next Skill Isn't More Code, It's Judgment
Forward-deployed engineers are having a moment. In a recent McKinsey Explainer, Holger Hürtgen, Martin Harrysson, and Nick Robson describe the FDE as an engineer embedded in a client's own systems and data, building alongside the people who'll use the result. The reason is that off-the-shelf AI rarely fits, and value only shows up when someone tailors the tool to the problem and helps the team change how it works. As Hürtgen puts it, "FDEs have to be able to make judgment calls." Without that judgment, he warns, companies end up implementing AI for its own sake: token spend climbs while the bottom line stays flat.
Research from MIT CISR points the same way. In their August 2026 briefing, Woerner, van der Meulen, and colleagues open one proposition with a line every engineering team should hear:
"No one will remember how many AI agents you built."
Prototypes are cheap, but the effort they save is only deferred until production, and the advantage goes to teams that ship with provable outcomes. The same briefing names a quieter risk: the skills crisis is less about AI training than about losing the domain expertise needed to challenge polished AI output. The authors' advice is to keep people doing the complex work that builds judgment.
For an FDE, much of that judgment comes down to decision rights. The AI Decision Matrix from Ina Sebastian, Peter Weill, Thomas Haskamp, and Jan vom Brocke sorts decisions by ambiguity and risk into four types: routine, consequential, exploratory, and strategic. Each calls for a different split between humans and AI in framing the decision, acting on it, and learning from the outcome.
Harrysson describes an FDE's first weeks as sitting beside the people doing the work, learning who decides, what matters, and which data counts, much of it undocumented. Being technical isn't enough, he adds, if you can't uncover how decisions really get made. That's the practice Maru ClientLab is built for. Engineers already know how to build the agent. In ClientLab they rehearse the forward-deployed part: interviewing stakeholders, earning the data, defending the ROI, and deciding where a human stays in the loop, before the stakes are real.
References
H. Hürtgen, M. Harrysson, and N. Robson, "What Is a Forward-Deployed Engineer?," McKinsey Explainers, McKinsey & Company, October 16, 2026, https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-a-forward-deployed-engineer
S. L. Woerner, P. Weill, N. van der Meulen, I. M. Sebastian, M. Mocker, G. Benedict, and P. Reynolds, "AI Value Creation: Five Provocative Propositions," MIT CISR Research Briefing, Vol. XXVI, No. 8, August 2026, https://cisr.mit.edu/publication/2026_0801_AIValuePropositions_Woerner
I. M. Sebastian, P. Weill, T. Haskamp, and J. vom Brocke, "Designing Decision Rights for AI," MIT CISR Research Briefing, Vol. XXVI, No. 6, June 2026, https://cisr.mit.edu/publication/2026_0601_AIDecisionMatrix_SebastianWeillHaskampVomBrocke