BTECH research seminar

Navigating specialization and flexibility in machine learning entrepreneurship

Early stage, technology-driven ventures must navigate a trade-off between specialization and flexibility in contexts where the optimal path is unknown. Much of the digital entrepreneurship literature has construed digital technologies as highly malleable, affording founders with the ability to adapt and redeploy the technology to other ends as needed. However, recent work on the ontology of digital artifacts proposes a greater role for external actors – such as users – in shaping technologies through their interaction with it, limiting the agency of founders. To explore the interplay of actors and technology in shaping the growth trajectory of the firm, we conduct an inductive study of six ventures commercializing machine-learning capabilities. We show that the incorporation of training data builds an inadvertent rigidity into the machine learning capability, while simultaneously limiting the pool of actors who interact with the technology. This results in a degree of specialization that is hard to reverse, even in circumstances where the interests of founders and investors demand a return to a flexible approach. Our findings advance theory in digital entrepreneurship by introducing the possibility of categories of digital artifacts that do not share the malleability of digital technologies broadly.