Before AI writes the deal memo, can you defend the numbers?
The useful answer is one the investment team can trace back to the source.

An offering memorandum comes in as a PDF. Someone pulls figures into a comp sheet, checks an email for an updated assumption, and moves the numbers into a model. By the time the deal reaches a meeting, the team has an answer, but tracing that answer back to the right source can take another round of work.
Altus Group recently wrote about this problem in commercial real estate. A rent roll can move from a PDF to a spreadsheet and then into a valuation model. Each handoff creates another opportunity for a number or version to change. AI can produce a polished response from those files, but the response is only useful if the team can check how it got there.
This is close to what we have seen while working with a real estate investment team. Useful deal information lived across offering memoranda, comp sheets, shared files, and email. The team did not need one more place to ask questions. It needed a repeatable way to collect the relevant fields, put them into a shared working sheet, and review them before using them in the next analysis. We helped demonstrate that workflow with the team’s own deal material. Its time savings still need to be measured.
For me, that is the difference between asking AI to write a deal memo and asking it to support the work behind one. Which file is the current source? Which rent or expense figure was extracted? Does it match the workbook? What needs an analyst’s attention before the number moves into a recommendation?
AI can help with extraction, comparison, and preparing a first pass. The investment team still owns the assumptions and the decision. In fact, getting some of the copying and searching out of the way can give people more time to challenge the deal, talk through risk, and notice what the numbers do not explain.
At Ruby, we map those handoffs with the people who do the work, then build training around the tools and review steps they actually need. A useful underwriting workflow should make it easier to find the source of a number, not just faster to put that number on a page.
We will walk through how to find and improve workflows like this at our free Beyond the Prompt lunch and learn on November 19.
Sources & further reading
Start with the work. Build the right system around it.
We help teams understand where AI is useful, choose the right tools for the job, and build the confidence to keep improving.
Book a Call