Yu Fang is Co-Founder & CTO at Sonatus, specializing in distributed systems and AI-enabled vehicle architectures.
getty
Modern products behave less like applications and more like ecosystems. A single customer experience spans embedded controllers, cloud services, mobile apps, supplier components and real-time data pipelines. As challenging as it can be to build these systems, the bigger challenge for engineering teams today is understanding them.
Every organization is facing the same problem: engineering teams don’t lack data. They lack context for that data.
Critical information lives everywhere and nowhere at once. Design documents, test logs, manufacturing notes, field reports, supplier data, customer tickets—each holds a fragment of the truth. But no engineer, no matter how experienced, can stitch these fragments together fast enough to diagnose issues confidently. Systems generate more context in a week than a team can reasonably review.
This is the context trap: the information exists, but it’s too fragmented, too unstructured or too distributed for any human to gather manually.
For decades, engineering teams relied on structured signals like logs, metrics or sensor data to understand failures. These signals are essential, but they represent only the visible tip of the iceberg. The real root causes often live in the submerged mass below the waterline:
• A configuration change buried in a release document
• A pattern across historical incidents
• A dealership report describing behavior outside formal tests
Traditional tools couldn’t unify these signals because they spanned formats, systems and organizations. The result was that teams chased symptoms instead of causes, and the costs showed up as recalls, downtime or customer frustration.
Generative AI seems like a breakthrough in engineering automation, but even more value lies in its ability to synthesize information that has never lived in one place or one format.
Generative AI can ingest and connect information across previously isolated domains such as logs and sensor data, design documents, test results, field reports, manufacturing records and more. It can read what humans can’t, remember what humans forget and connect dots that humans don’t see.
Generative AI’s ability to retrieve all this information regardless of format and location is underappreciated.
With the help of appropriate tools, generative AI can almost instantly surface relevant specs, historical failures or obscure manufacturing notes from hundreds of documents—information a new engineer might not know exists, or an experienced one would spend days looking for.
For the first time, engineering organizations can zero into the right context at scale.
When AI can "see" across the entire system, "build" the knowledge graph and "reason" on top of it, engineering work shifts from reactive troubleshooting to proactive understanding.
AI can identify patterns across incidents, correlate them with configuration changes and propose likely root causes.
Engineers can evaluate AI-generated hypotheses against outliers, compliance requirements and safety constraints.
Institutional knowledge, once trapped in people’s heads or scattered across documents, becomes accessible on demand.
AI can connect a symptom in the field to a manufacturing variation to a design assumption made years earlier.
This isn’t automation replacing engineering expertise. It’s complementing what human expertise can see and solve.
To take advantage of this shift, organizations must rethink how they capture and structure engineering context. Generative AI is powerful, but only if the underlying "signals" are accessible and interpretable.
Three changes matter most:
In an AI-assisted world, logs must capture decision context, not just events. Engineers need to know why a system behaved in a certain way, not just what it did. Logging becomes an intentional part of system design, not an afterthought.
Design docs, test plans and field reports are no longer created for humans alone. They are inputs to AI systems that will analyze, correlate and retrieve them. Their clarity and completeness will directly affect the quality of diagnostics.
AI can only synthesize the data it can access. Leaders must make logs, test results, field reports and documentation available to systems that can correlate them, with appropriate governance and security.
Engineering leaders must treat AI not as a black box, but as a partner in investigation. That requires cultural and technical shifts.
The near-term goal is to hunt faster; the long-term goal is to eliminate the hunt altogether.
AI-generated insights must be evaluated with the same rigor applied to any unfamiliar code or hypothesis.
Some of the most valuable signals live outside traditional telemetry. Treat these sources as core diagnostic input, not peripheral artifacts.
Organizations that embrace this shift will diagnose issues faster, ship more reliable products and build systems that improve over time. Those that don’t will find themselves trapped in reactive firefighting mode, unable to keep pace with the complexity they have created.
Generative AI doesn’t replace engineering judgment. Rather, it finally gives engineers the context they’ve always needed but could never achieve manually. The next era of engineering will belong to the teams that can leverage AI to find, relate and reason on all the relevant "signals" available throughout the system.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。