Audit your deal week before you delegate a line of it
Lesson 1 of 5 · The deal week
By the end you can turn one week of underwriting, calls and memo work into a table that says what an assistant drafts, what it checks, and what never leaves your desk.
1 · Learn the move · Template pattern
The first AI job on a deal desk is usually chosen by what demos well: a chatbot on the pipeline, an auto-generated pitch. Choose it from your own timesheet. Write last week task by task with hours and whether it repeats, then sort every line. DRAFT: the model writes a first version from inputs you supply (the memo narrative from a verified rent roll, the LOI cover note, the weekly pipeline update). CHECK: it tests one thing against a record you supply (a T-12 against the operating statement, lease dates against the estoppel, a comp set against the source). YOURS: the cap rate you underwrite to, the value opinion, anything that will be shown to an investor, anything a lender relies on. That third column is not the tool's limit. It is the line that keeps your name off a complaint. The best first job repeats weekly, eats the most hours, and sits entirely in DRAFT or CHECK.
Here is my last week, task by task, with rough hours: [paste]. Sort it into a table: task | hours | repeats weekly? | DRAFT / CHECK / YOURS | one line why. Value opinions, return assumptions and anything an investor or lender will rely on go in YOURS. Then name the one repeating task with the most hours that is fully DRAFT or CHECK. If my notes do not say how long something took, write [NEEDS: hours].
2 · Your turn. You write the prompt
Friday, 5 pm. This week you rebuilt a rent roll from three PDFs, wrote two LOIs, sat through a lender call about DSCR covenants, updated the pipeline deck twice, pulled comps for a BOV, and spent most of Wednesday reconciling a T-12 that did not tie to the bank statements. The partners want 'an AI plan' by Monday. You want the eight hours back first.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.