McKinsey article on agentic AI in real estate operating models. Argues that current AI deployments sit adjacent to workflows as small peripheral tools and miss the structural redesign needed for agents to deliver real value. Lays out a five-layer architecture (data, model, orchestration, action, control) and the prerequisite organisational work: workflow redesign at the domain level, governance, structured data, role definitions, and explicit ownership of the learning loop. > Without the right architecture, agents can't work together... organizations end up with impressive demos that cannot scale. ^mckinsey-on-demos-not-scaling > Stop asking, "What use cases can we pilot?" and start asking, "Which workflows should we redesign so the software is allowed to do the work, with appropriate controls?" ^mckinsey-on-workflows-vs-use-cases > Who owns the learning loop - the owner, the property manager, the software vendor, or the services provider? ^mckinsey-on-owning-learning-loop