Mary Shaw, 2015

AI-generated editorial portrait of Mary Shaw in a university software engineering setting

“The greatest need for engineering is in the most critical applications.”

Mary Shaw delivered this judgment in her keynote “Progress Toward an Engineering Discipline of Software” at GOTO Amsterdam in 2015. She was asking whether software development had matured from craft and commercial practice into an engineering discipline able to make dependable predictions about the systems it creates. Her answer was deliberately conditional. Not every small program needs the same rigor, but software that operates unattended or controls consequential processes deserves methods proportionate to the harm a failure could cause. In that setting, “engineering” means more than professional identity: it means making tradeoffs explicit, drawing on codified knowledge, and analyzing a design before people are forced to discover its weaknesses in use.

Shaw’s wider career explains the emphasis. At Carnegie Mellon University, she helped establish software architecture as the study of a system’s large-scale structure—its components, the connections among them, and the design choices that bridge requirements and code. Architecture makes knowledge visible so teams can compare alternatives instead of treating every project as an unprecedented act of invention. In her 2015 talk, Shaw used examples such as medical implants, nuclear-safety systems, self-driving vehicles, and automated trading. These systems differ technically, but they share a crucial feature: when something goes wrong, the opportunity for a person to notice and intervene may be brief or nonexistent.

The quotation does not say that ordinary software may be careless, or that formal methods alone can guarantee safety. “Greatest need” establishes a priority, not an exemption. Engineering judgment still works under limits of time, evidence, and resources, and different risks call for different combinations of testing, review, modeling, monitoring, and operational safeguards. Nor is criticality determined only by dramatic physical danger. A failure in financial, communications, or public-service software can also deny access, expose information, or impose costs at scale. The important question is not whether a product looks sophisticated; it is what depends on it, how failure can propagate, and whether recovery is possible before damage becomes irreversible.

That distinction is increasingly useful as software is embedded in infrastructure and delegated more decisions. Cloud services depend on long chains of components, connected products act in the physical world, and machine-learning systems can produce outputs faster than people can inspect them. Shaw’s line offers a practical test for technology leaders: the more autonomous the system and the more serious its failure modes, the stronger the evidence its designers should demand. Clear requirements, traceable decisions, independent checks, and rehearsed recovery plans are not bureaucratic decoration. They are how an organization earns confidence that a critical system will behave as intended—and how it prepares honestly for the cases in which it will not.

Mary Shaw presented the line during her 2015 GOTO Amsterdam keynote, which examined how software practice evolves from craft toward an engineering discipline grounded in shared, codified knowledge.

She focused especially on automated and unattended systems, where a failure may cause serious harm before a person has time or information to intervene.

“Greatest need” is a risk-based claim: engineering rigor should rise with a system’s autonomy, scale, and consequences, not merely with its novelty or commercial profile.

The quote does not promise certainty. Testing, analysis, review, monitoring, and recovery planning provide different kinds of evidence, and responsible teams choose them according to the failures they must prevent or contain.

Critical software now mediates medicine, mobility, finance, communications, public services, and infrastructure, often through complex chains of vendors, libraries, and cloud systems.

Shaw’s principle asks organizations to match confidence claims with evidence and to prepare for failure before an automated system makes human correction too late.

Explore more "Quotes of The Day"

Discover more notable quotes from influential voices across politics, science, business, technology, sports, and culture. Each quote offers insight into how ideas, beliefs, and decisions shape the world around us.