Practice Briefing: From automation to delegation – governing generative AI in property investment

Purpose Most real estate organisations approach artificial intelligence with the question “what can we automate?” This paper argues that the framing is wrong for generative AI, whose highest-value uses in property investment are not the automation of known processes but the delegation of judgement-laden work – and that the industry's default safeguard, “a human in the loop”, is not a governance strategy. Design/methodology/approach A practice briefing combining the management literature on delegation and recent evidence on human–AI complementarity with the authors' experience of three years of executive education and advisory work with senior real estate professionals. Findings Generative AI creates most value where neither the process nor the output can be fully specified in advance – precisely where automation-era governance fails. Uniform human review predictably degrades into rubber-stamping as volumes grow. Effective governance separates two questions: what kind of work is being transferred, and how much the outcome matters. Practical implications A two-axis classification (is the process known? Is the output verifiable?) Sorts any task into one of four working modes – automate, assist, structure, delegate – and an accountability ladder scales oversight from logging to full human decision-making. Investment teams can apply the framework immediately by classifying tasks rather than jobs and reallocating scarce review capacity to where consequence is genuine. Originality/value The paper reframes AI adoption as a delegation-design problem continuous with management skills the industry already possesses, separates task classification from governance intensity and names the failure mode – accountability theatre – that uniform human review creates.