Skip to content
Episodes The Framework The Menu About Work with Reu The Diagnostic Book Reu →
← The Menu ~ station 05 of 06 · the finance station

The Finance Station

The Finance Station turns raw numbers into decisions and keeps the books from running the business. Most operators have AI doing categorization and AP triage already. Those are the easy plates. The hard plates are variance commentary, the close narrative, the board package. The work where numbers become a story leadership can act on. The Chef's habit problem on Finance isn't AI gets the math wrong. It's we never wrote down what a board member needs to see in two minutes.

What this station holds

Finance is a two-tier operation. The first tier is plumbing: invoices hitting the right categories, aging AR getting follow-ups, transaction data flowing into the ledger without a human touch. The second tier is narrative. Variance explained in plain English. A cash forecast with assumptions called out. A board story that makes sense in 120 seconds. A case for capital allocation that doesn’t require a spreadsheet degree to understand.

The finance team today spends their week on both. Hours on categorization and invoice triage. Hours building close narratives that should take 45 minutes but take 4 days because the CFO’s voice is hard to find in software. Scenario modeling that lives in Excel and never gets presented because the assumptions are buried. Board packages assembled from five different places, missing context, reading generic.

The Finance Station is built to reclaim the time on the easy work and raise the bar on the hard work. It reads the transaction feed and tags it right. It validates invoices against terms and pulls them into the approval queue. It forecasts 13 weeks of cash with assumptions visible. It drafts variance commentary in language a board member actually gets. It models a hypothesis and surfaces what the numbers say.

When the station runs well, the close happens a week faster. Variance gets explained before anyone asks the question. The board package reads like the CEO wrote it because the station learned from three examples the CFO actually drafted. Pricing decisions surface the places margin is leaking instead of the CFO discovering them at year-end. Capital asks come with a structured argument instead of a gut call that takes three meetings to defend.

The dependency property within the station

Pull the Chef and the station generates correct output but the CFO never knows what questions to ask next. Pull the Sous Chef and invoices stack in the wrong approval queues. One stays open for two weeks because nobody routed it to the right person.

Pull any single early dish (expense categorization, AP triage, AR follow-up) and the station fragments. Categorization fails without confidence thresholds that route edge cases to human eyes. AR follow-ups land generic without examples of house voice. Cash forecasts diverge from reality when the assumptions aren’t written down upfront.

The difference from other stations is the weight on voice and judgment at the back end. Dishes 1-3 run on structured input and recoverable output. Transactions either go to the right account or they don’t. Dishes 4-6 cross a threshold where output style matters. The CFO’s voice on variance commentary can’t be faked. The close narrative needs to sound like the company thinks, not like bookkeeping software thinks. Dishes 7-10 are judgment work. Pricing analysis surfaces patterns but the CFO decides what to act on. Scenario modeling is a thinking partner, not an analyst replacement. The Chef stays in the loop on every plate.

The dish order

The first three dishes run on what already sits in your business. Transaction feed. Invoice archive. Past AR follow-ups. The training set is already written. You just haven’t pointed the station at it.

The middle four cross a threshold. Variance commentary needs CFO voice wired in before deployment. Cash forecasting needs a scoreboard to measure accuracy. Close narratives need examples from your hand before the station ever drafts one. Guardrails have to be defined before deployment, not after.

The last three are where most operators stall. Pricing analysis surfaces the patterns but doesn’t decide. Scenario modeling is the Chef thinking with the station, not the station thinking for the Chef. Board packages need the CEO’s voice or they get read once and forgotten. These don’t live at the pass because the CFO is busy. They live at the pass because the recipe gets sharper when the Chef is tasting.

  1. Expense Categorization. Every transaction tagged to the right account, project, and tax category. At the moment it hits. See expense-categorization.md.
  2. Invoice and AP Triage. Inbound invoices read, validated against PO or contract, queued for approval with the right context. See invoice-and-ap-triage.md.
  3. AR Follow-Up Drafting. When invoices age past terms, draft the right collection email. Tone calibrated to the relationship. See ar-follow-up-drafting.md.
  4. Variance Commentary. The actual vs. budget delta, written up in plain English. What moved, why, what to watch. See variance-commentary.md.
  5. Cash Flow Forecast Drafting. Take the AR aging, the AP queue, the recurring obligations. Return the next 13 weeks of cash with assumptions called out. See cash-flow-forecast-drafting.md.
  6. Financial Close Narrative. Monthly close summary, written for the leadership team. Not the accounting team. See financial-close-narrative.md.
  7. Pricing Analysis. Pull margins by customer, segment, or product. Surface the three places pricing is leaking. See pricing-analysis.md.
  8. Scenario Modeling Support. Given a hypothesis, draft the model and the narrative. Station is a thinking partner, not analyst replacement. See scenario-modeling-support.md.
  9. Board Package Drafting. The quarterly story for the board. Performance, risks, asks. Assembled from the system, written in the CEO’s voice. See board-package-drafting.md.
  10. Capital Allocation Framing. When a real decision is on the table. Station drafts the case for and against, grounded in actual numbers. See capital-allocation-framing.md.

What breaks this station

Categorization without guardrails. The chart of accounts gets written down but the station never sees the confidence thresholds. When a transaction is ambiguous, confidence should be below 85 percent. Route to human eyes. Without that rule, wrong categorizations stay in the ledger.

AP triage without context. The station reads “invoice received” but never sees the PO or the contract. It approves a duplicate. It queues a bill to the wrong person. Context plumbing is the first thing to get right.

AR follow-up defaulting to aggressive. Every aged invoice gets the same tone. The station doesn’t know that this customer is at-risk, this one is a good account with a slow process, this one needs a gentler touch. House style on tone and relationship context is load-bearing.

Variance without a destination. The station drafts variance commentary. The CFO looks at it. They rewrite half of it because the station missed the why, missed the pattern, missed what the team should do about it. The recipe never gets sharper because the feedback loop died.

Forecast assumptions hidden. The station says “cash is +$200K next month” without writing down the assumptions. When the forecast misses, the CFO can’t trace why. When decisions are made on the forecast, the assumptions never made it into the decision.

Close narrative that sounds like QuickBooks. The station learned to write “Gross Margin improved due to higher transaction volume.” That’s technically true. It’s also hollow. The CFO reads it, rewrites it, the station never learns the difference between accurate and meaningful.

The Monday Move for this station

Pick this month’s close. Use the station to categorize one week’s transactions. You review the categories. Correct any misses. That’s your confidence threshold now. Use the station to draft variance commentary on one P&L line. You write one by hand. Compare. Write the first three variance commentaries yourself. Show the station. That’s your house standard. The station is now running. Finance is at the pass.


Original framework. Distilled from client work. Companion to The Station Plan and The Professional Recipe.

The dishes on this station

01

Expense Categorization

Every transaction that hits the ledger should be tagged to the right account, project, and tax category the moment it posts. A human…

02

Invoice and AP Triage

Every inbound invoice that shows up should be validated against what you actually ordered, matched to the right PO or contract, checked for…

03

AR Follow-Up Drafting

When an invoice ages past terms, the next thing that happens is a follow-up. The follow-up is where tone matters more than most operators…

04

Variance Commentary

The actual vs. budget delta needs to be explained in plain English. What moved, why, what to watch. The leverage isn't the math. The math is…

05

Cash Flow Forecast Drafting

Take the AR aging, the AP queue, the recurring obligations, and return the next 13 weeks of cash with assumptions called out. Most operators…

06

Financial Close Narrative

The monthly close summary, written for the leadership team, not the accounting team. The difference between a close narrative the CEO reads…

07

Pricing Analysis

Pull margins by customer, segment, or product. Surface the three places pricing is leaking. The station surfaces the patterns. The CFO…

08

Scenario Modeling Support

When the CFO needs to model "what if we raised pricing by 5 percent" or "what if churn went from 3 percent to 5 percent," the station runs…

09

Board Package Drafting

Every board meeting needs a financial narrative written in the voice of the CEO who built the company, not the voice of a bookkeeper…

10

Capital Allocation Framing

When the Chef needs to decide where cash goes next, the station frames the options. Which problem to solve first. What the tradeoff is. What…

~ next on the line

The Research Station

The Research Station turns information into the operator's edge without becoming a research organization you don't need. Most operators think research means stalking competitors and hoping a pattern emerges. The real work is narrower. Define the signal. Define the cadence. Define the format. Let the station run. The Chef's habit problem on Research isn't AI gives me too much information. It's we never wrote down what counts as signal versus noise. Without that definition, the station drowns in facts. With one, it produces a 200-word morning briefing that actually changes Monday's move.

Walk to the next station →
← The Operations Station ← Back to The Menu The architecture behind the kitchen →
AI in Crayon

AI strategy. Translated for leaders.

New episode Tuesday, Thursday, Friday.

~ est. 2026

The Show

All Episodes Latest

The Framework

The whole system The Station Plan The Four D's The Prep List The Professional Recipe Quality Control The AI Audit

Reu

About Book Reu Local Nerds Email
© 2026 AI in Crayon Built by Local Nerds