tp.alptransfer pricing at arm's lengthDoe & Roe LLC

Where AI Helps

Where a model saves the team real time — and the guard that keeps each use inside the spine: a suggestion, quoting its source, decided by a person.

WhereStatusWhat it does, and the guard
Reading agreementsBuilt — switched on with the firm's OpenRouter key (off on this deployment)A model extracts parties, rates, payment terms and risk, each with its clause and a quote; every quote is checked against the text word for word; the result is an item to review. Through OpenRouter.
Screening comparablesRules-based today; a model nextRead each company's business description from the database export and propose accept or reject with the sentence relied on — the slowest manual step in a study. Already a labelled suggestion; a model would read the descriptions a keyword list cannot.
The functional analysis from interview notesProposedTurn the team's interview notes into answers to the questionnaire, each answer quoting the note it came from, for the team to accept or change.
Drafting Local File sectionsDrafted by rules todayA model writing the narrative from the items, quoting them, never adding a figure that is not in an item; the section is then signed off like any other, and the file exports as final only when every section is confirmed.
Roll-forwardProposedRead last year's Local File and this year's figures and list what must change in the text, section by section.
Ask The RulesBuilt — switched on with the firm's OpenRouter key (off on this deployment)Answers "does a cash pool count?" from the rules library only: every answer cites the rules it rests on, unknown citations are dropped, an answer citing nothing is refused, and what the library does not cover is said to be not in it.
Information requestsProposedDraft the chaser to the client for what is overdue, in English or Greek, for a person to send.
Never for client data on this demonstration. Inside the firm, the same uses run on the firm's own approved model environment, with the same guards.