The Specific Claim General Motors Is Making

General Motors made a set of precise public commitments this week: smarter vehicles, faster over-the-air software updates, and autonomous features capable of handling mundane tasks like dry-cleaning runs. The framing from GM's leadership is consistent and deliberate. This is not a car company that happens to use software. This is, in their telling, a software-defined technology company that happens to currently sell trucks and SUVs. The gap between those two self-descriptions is not a branding question. It is an organizational competence question, and it is worth taking seriously.

What "Software-Defined" Actually Requires From an Organization

A hardware manufacturer announcing that it is becoming a software company is making a claim about internal coordination capacity, not just product roadmap. Software-defined vehicles require continuous update cycles, modular architecture decisions, and real-time responsiveness to algorithmic feedback from vehicle fleets. These are not skills that transfer automatically from manufacturing excellence. The organizational routines that make GM effective at building physical vehicles at scale - supplier coordination, quality inspection, logistics sequencing - are largely orthogonal to the routines required for software iteration and AI deployment governance.

This is where organizational theory becomes clarifying rather than decorative. Rahman (2021) documented how algorithmic management systems create "invisible cages" that constrain worker behavior in ways that even organizational leadership often does not fully perceive. The relevant insight here is directional: organizations do not simply adopt AI systems and then operate them. The AI systems begin reshaping what the organization can perceive, respond to, and coordinate around. For GM, deploying AI across its vehicle fleet means the organization must develop new sensing capacities for a class of operational signals it has never historically needed to monitor.

The Competence Transfer Problem GM Is Not Advertising

The variance puzzle at the center of my dissertation research applies here in an underappreciated way. When platform workers with identical access produce dramatically different outcomes, the explanation is not differential effort or raw ability. It is differential schema quality - specifically, whether a worker has an accurate structural model of how the algorithmic environment actually functions, rather than a folk theory assembled from surface observations (Kellogg, Valentine, & Christin, 2020). The same logic applies to organizations attempting platform transitions.

GM's engineers understand combustion systems and supply chains with genuine depth. That is procedural expertise, specific to a particular technical and organizational context. Hatano and Inagaki (1986) drew the foundational distinction between routine expertise - the capacity to execute known procedures effectively - and adaptive expertise - the capacity to generate novel responses when the procedural environment changes. A software-defined vehicle company operating AI systems across a live fleet requires the second kind. There is no evidence that routine expertise in vehicle manufacturing transfers to adaptive expertise in algorithmic fleet coordination, and there is considerable theoretical reason to expect it does not.

The Governance Problem That Is Actually New Here

The autonomous dry-cleaning run detail in GM's announcement deserves more analytical attention than it typically receives in coverage. This is not a metaphor. It is a specific claim about delegating decision-making authority over vehicle routing, timing, and environmental response to an algorithmic system operating without real-time human oversight. Sundar (2020) identified the rise of machine agency as a distinct communication and coordination problem: when machines act as agents rather than tools, the accountability structures that organizations rely on become structurally ambiguous.

Who is responsible when an autonomous vehicle makes a routing decision that produces a negative outcome? The answer is not obvious inside a traditional automotive organizational hierarchy. GM's announcement implies a governance architecture that does not yet exist publicly in any documented form. Hancock, Naaman, and Levy (2020) framed AI-mediated communication as a category that requires new theoretical scaffolding precisely because existing frameworks assume human agents at the endpoints of consequential decisions. Autonomous vehicle operation at the task level breaks that assumption at scale.

Why This Matters Beyond GM

GM's pivot is a case study in a broader organizational phenomenon: the awareness-capability gap applied at the institutional level. GM's leadership is clearly aware that AI integration is structurally transformative. That awareness, however, does not automatically generate the internal coordination capacity to execute the transformation. Knowing that software-defined architectures require different organizational routines is not the same as having those routines. The company's public framing is sophisticated. The harder question is whether the internal organizational schema matches the external claim, and that question will not be answered by a product announcement.

References: Hatano, G., & Inagaki, K. (1986). Two courses of expertise. In H. Stevenson, H. Azuma, & K. Hakuta (Eds.), Child development and education in Japan. Freeman. Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366-410. Hancock, J. T., Naaman, M., & Levy, K. (2020). AI-mediated communication: Definition, research agenda, and ethical considerations. Journal of Computer-Mediated Communication, 25(1), 89-100. Rahman, H. A. (2021). The invisible cage: Workers' reactivity to opaque algorithmic evaluations. Administrative Science Quarterly, 66(4), 945-988. Sundar, S. S. (2020). Rise of machine agency: A framework for studying the psychology of human-AI interaction. Journal of Computer-Mediated Communication, 25(1), 74-88.