Run one listing pricing question on a fast model and then a reasoning model to get a defensible price and the seller rebuttal

In Agent Intelligence issue #75, Matt Spangler takes a real listing appointment where the sellers want a number the comps don't support. The agent pastes the address, beds, baths, square footage, the sellers' number, and three closest comps (or an existing comparative market analysis) into ChatGPT or Claude and asks for a recommended price, the two strongest data points, the exact objection the sellers will raise, and a word-for-word answer. The agent runs the same prompt twice: first on the default model, then with the reasoning model and high effort turned on. Spangler reports the reasoning run weighs the spread between the comps and catches that the sellers are anchored high, while the default run gives a reasonable-sounding number.

Compliance note: The recommended price is a draft talking point, not an appraisal; confirm every comp in the MLS and keep the final list price the seller's decision. Leave seller names and personal details out of the prompt.

You are my listing strategist. Analyze [address], [beds/baths/sqft]. The sellers want to list at [$their number]. My three closest comps are [comp 1: price, size, condition, distance], [comp 2], [comp 3]. Give me the price you would recommend, the two strongest data points behind it, the exact objection these sellers will raise, and word for word how I answer it. Be specific to these comps, not generic.

← All use cases