Systems Thinking Is Vital in the Age of AI

Sufian Aldogom, an IEEE member, research affiliate at MIT and senior leader at Google in Sunnyvale, Calif., has published an argument that systems thinking should sit at the center of engineering practice as AI tools spread through design work. The piece is a position essay rather than an announcement of a product or policy change; it lays out how Aldogom believes engineering roles and professional institutions should adapt.
His starting scenario is a workflow many teams will recognize. An engineer frames a problem, then hands it to an AI tool to search research and technical standards, propose design options, run simulations and compare outcomes. The engineer then takes those results to specialists who can judge what the model cannot: whether a design can be built safely, whether it fits a particular setting, and how it affects the people who use it. The decision on whether and how to proceed stays with humans.
The shift Aldogom describes is one of scope. Work that once required several people with different technical backgrounds — literature review, initial modeling, option generation — can now be done earlier by one engineer working with AI. That moves the engineer's job from executing each technical step toward orchestrating people, machines and specialist knowledge around the problem.
He is careful not to claim that AI dissolves disciplinary expertise. Electrical, mechanical, civil and software engineering each take years to master, and AI does not remove that requirement. What changes, in his account, is how cheaply an engineer can explore territory outside their own field. His example is a software engineer on a medical device: AI can help review biomedical literature, decode unfamiliar terminology, analyze data and build a first model, but it does not turn that engineer into a physician. The benefit is asking sharper questions before the real experts are brought in. The same pattern holds in energy, transportation and manufacturing, where a software problem quickly pulls in hardware, safety, economics, human behavior and regulation.
From there the essay turns to systems thinking — looking at the whole problem, how parts interact, where constraints and failure points sit, and which trade-offs are acceptable. AI adds a layer to that discipline. Engineers must now decide not just how technical components fit together but which work goes to a machine, what a person must verify, when a specialist should take the decision, and when the problem itself needs reframing. Aldogom's formulation is that an engineer's value lies not in knowing everything but in framing the problem, asking the right questions, testing the answers, and knowing when another expert should lead.
He stresses that engineering problems rarely stay still. New findings can shift objectives, expose hidden limits or invalidate the original premise, so effective engineers revise scope as their understanding of the system deepens rather than clinging to initial specifications. Engineering, on this view, is as much a disciplined way of thinking as a body of knowledge.
The essay is pointed about AI's failure modes: confident answers built on weak assumptions, code that passes one test and fails the next, summaries that miss which evidence is strongest, and designs optimized for the wrong objective. Aldogom's division of labor is that machines suit speed, scale, search, pattern recognition and data generation, while people are needed for context, values, responsibility and judgment. He sketches a medical sensor project where AI proposes several designs and predicts performance, an engineer rejects the ones that violate technical requirements, a clinician judges whether the sensor is usable in practice, and patients and practitioners surface issues nobody upstream considered. The answer comes from combining those perspectives, and the engineer increasingly acts as a coordinator of knowledge alongside being a builder.
This is where he sees a role for IEEE. Its societies already span technologies, engineering education and technical standards, which he says positions the organization to support cross-specialty work without devaluing specialization. He proposes more communities organized around problems that need several fields — trustworthy clinical AI, resilient infrastructure, energy reliability, scientific discovery and safe autonomous systems are his examples — so an engineer can start in a home discipline and pull in others as a project demands.
He also wants practical standards for human-machine collaboration: guidance on when AI-generated work must be checked, how uncertainty should be documented, when a specialist must decide, and who is accountable when output mixes human judgment and machine generation. Leaving those questions to each company or project manager, he argues, produces inconsistent practice that does not travel with engineers between jobs, industries and fields. A professional body, in his view, can supply shared expectations.
His closing definition is of the "unbound" engineer: someone trained to understand complex systems, define and redefine problems, assemble the knowledge and tools to solve them, test whether solutions work, and take responsibility for the result. Disciplines matter more, not less, because depth is what lets an engineer judge whether an answer is sound, while AI makes the connections between fields easier to traverse. The core, he writes, stays human — choosing which problems are worth solving, what evidence is sufficient, how the pieces fit, and what to do under uncertainty.
For developers and product teams, the practical read is that AI-assisted engineering raises process questions as much as tooling ones. Teams building AI features into design, simulation or coding workflows may find that reviewers, auditors and customers start asking who verified machine output and who signed off — the kind of question standards work is meant to standardize. Organizations that treat AI as an accelerator for exploration while keeping explicit human checkpoints, documented uncertainty and clear ownership of final decisions will be better placed if such expectations harden into formal guidance. That is inference from the essay's argument rather than a reported development, but it follows directly from the accountability gaps Aldogom identifies.
Why it matters: Engineers and the teams that employ them are being pushed from executing technical steps toward coordinating people, AI tools and specialists, which changes hiring, review and sign-off practices. If professional bodies do produce shared standards for checking AI-generated work and documenting uncertainty, product teams may eventually face common expectations instead of inventing their own. The essay signals that debate is moving from whether AI belongs in engineering to who is accountable when it is wrong.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Modern engineering projects increasingly combine human judgment with machine capability. An engineer can define a problem, then ask an artificial intelligence tool to search relevant research and technical standards, generate design options, test them in simulation, and compare the results. The engineer can then bring the results to specialists who understand the parts AI cannot judge on its own, such as whether a design can be built safely, whether it makes sense in a particular setting, or how it might affect its