What to build first.

When a team gets access to something that can build software from a sentence, the instinct is to reach for the most ambitious thing on the wish list. The firm-wide portfolio system. The model that finally replaces the twelve-tab workbook. It is the right ambition and the wrong first move, because the biggest idea is also the one with the most opinions attached to it, and opinions are slow.

The apps that stick are smaller and more annoying. They are the tasks somebody does every week, by hand, slightly resentfully.

Four that tend to work

• The rent roll you keep re-formatting. Drop the file in, get a clean summary with unit mix, occupancy, loss to lease and expiries. It takes minutes and immediately saves an afternoon a month. See rent roll dashboards.
• The T-12 normalization step. Every broker exports differently, and someone always maps the line items by hand. Let it map itself and spot-check the result instead. See what a T-12 is if you need the primer.
• The screening pass on new deals. A quick, consistent first look so the same questions get asked of every deal that lands, before anyone spends real time on it.
• The status report nobody wants to write. The Monday pipeline update, the quarterly asset summary. Recurring, formatted, and perfect for an agent to run on a schedule. See agent automation.

The one that never works first

"Rebuild our entire underwriting model, exactly as it is." Not because it cannot be done, but because that model encodes a decade of house convention, and every reviewer will have a different note. Build it second, once the team already trusts the tool and knows how to steer it. By then the arguments are about assumptions rather than about whether any of this is real.

How to phrase the first ask

Brief it the way you would brief an analyst, in outcomes rather than instructions. "Watch my rent roll and flag units more than ten percent below market" is a better prompt than a description of a spreadsheet layout. You will get a working app with its own database, and you refine from there rather than starting from a blank page.

Two habits make the difference in week one:

• Use real data immediately. Sample data teaches you nothing about whether the output is trustworthy. Drag in an actual file, or start against the CRE data already loaded in your account.
• Share it early. An app one person uses is a personal shortcut. An app the team opens is a change in how the shop works.

Start with the annoying, weekly, unglamorous thing. Momentum comes from the second and third app, and you only get those if the first one actually saved somebody an afternoon.