How should organizations deploy AI as capabilities evolve—deciding what to build and what to buy, bringing together business experts, engineers, and data teams, and turning technical success into sustained adoption?
How should organizations deploy AI as capabilities evolve—deciding what to build and what to buy, bringing together business experts, engineers, and data teams, and turning technical success into sustained adoption?
In early 2023, Tony Kim, a Managing Director at BlackRock, became convinced that generative AI would reshape knowledge work. That included fundamental equity research, which AI had touched far less than the firm’s quantitative investing. Kim believed the greatest challenge would be getting investors, data teams, and AI engineers—with different cultures, roles, and vocabularies—to collaborate effectively. Working with BlackRock’s AI Labs, he assembled a small team to build BlackRock Asimov, an agentic platform that supports research and monitors investment theses. Investors on the team contributed as both domain experts and first users, shaping workflows, testing prototypes, and encoding their research methods. Engineers developed the agents, orchestration, and evaluation processes, while data specialists organized fragmented research into a shared, governed data layer. Together, they combined external models with proprietary capabilities while preserving flexibility to adopt new technologies. Publicly unveiled in June 2025, BlackRock Asimov met its technical objectives. By late 2025, its standardized applications had engaged roughly half of the intended users, raising the next challenge: how to build on that initial success while accommodating research practices and preferences beyond those of Kim’s own investment team.
Told in two parts through the perspectives of investors, AI engineers, and data leaders, the case examines AI deployment as a set of interdependent strategic, technical, and organizational choices. Students consider how leaders act on an emerging technology shift and revisit build-versus-buy decisions as models improve; how to balance cost, quality, latency, and control; and how to involve users throughout development and deployment. They recommend a path toward broader adoption. The second part explores BlackRock’s approach but also how investments in data, architecture, and governance enable new ways of working, with implications for incentives, competitive advantage, and professional roles.