Compare the combined workflow with credible alternatives. Measure the quality of the decision record and the expert effort it requires, using only information available at the time of each case.

A faster investment memo can still require more work downstream. A committee may need to reconstruct assumptions, resolve contradictory sources or question how a confident recommendation was reached. A useful test of AI in real estate investment therefore follows the whole task through to a reviewable decision record.

Hybrid intelligence means experts and AI working on the same decision, each doing what it does best, by design. The benefit is a proposition to test. It should not be inferred merely from having both a person and an AI tool involved.

Choose a decision-sized question

Start with a bounded task: identifying the assumptions that drive a recommendation, reviewing downside scenarios or preparing the evidence for a committee discussion. Define the input documents, the required output, the expert role and the point where the task ends.

Keep assumptions about the asset separate from facts in the source material. For example, an expectation about lease renewal should remain distinguishable from an agreed renewal. The test should examine whether uncertainty survives summarisation and is carried into the final reasons.

Compare three arrangements

Where practical, compare experts working without AI, AI producing an output on its own, and the combined workflow. The AI-only condition is a testing reference, not a proposal to delegate investment authority. Use comparable case difficulty and avoid giving one condition information that another does not receive.

A 2024 systematic review and meta-analysis by Vaccaro, Almaatouq and Malone found that human-AI combinations did not, on average, outperform the better of the two working alone. The research covered many tasks and is not evidence about any particular real estate practice. It supports asking for an explicit comparison instead of assuming that combination itself creates an advantage.

Define success before seeing the outputs

Agree criteria with the people who will use the work. Useful dimensions include the preservation of material assumptions, the quality of source support, treatment of conflicting evidence, clarity of downside reasoning and consistency of escalation. Include the ability to say that the evidence is insufficient.

Measure expert time across preparation, review, correction and discussion. Time released is capacity: it becomes useful when the firm decides how to deploy it. It should not automatically be reported as a cash saving.

Guard against hindsight and rehearsal

Past investment cases are valuable only if the test controls what participants know. Reconstruct the information available at the original decision point and separate it from later outcomes. A favourable eventual return does not, by itself, prove that the original reasoning was sound.

Reserve cases that were not used to develop the revised workflow. Where reviewers could remember a case, use separate case sets or a planned allocation that reduces that effect. Record the model version, prompts, source material and workflow rules so that changes in the setup remain visible.

Turn the result into an operating decision

Test, redesign, test again means the second comparison can support adoption, further revision or rejection. Record what improved, what did not, which cases remain difficult and what should trigger another review. A stronger decision process is the objective; a more impressive demonstration is not the measure.