The cash levers: collect what you're owed and cut spend with evidence, not vibes

Your team already has the ar-collections and vendor-spend playbooks — the worked recipes for running an AR aging and drafting reminders, and for pulling every subscription onto one page to spot what to cut. Each is a good move, done once. This module is the layer above the recipe. It’s where the two levers that actually move cash — collect what you’re already owed, and control what you spend — become a discipline: traced, evidenced, ranked, and graded on whether the numbers would survive a CFO putting them in front of the board.

It’s Module 4 of the certifiable Finance & Ops track, and it inherits everything you built in Module 1 — the categorization rulebook, the data-trust checklist, and the controls doc. It follows the monthly report (Module 2) and plan & forecast (Module 3): once you can close, report, and forecast, the cash levers are what you actually pull on those numbers. M1 gave you books you can trust; M2 and M3 turned them into a report and a forecast; M4 is where you move the money.

The fastest cash isn’t a new sale. It’s the cash you already earned that’s aging quietly in someone else’s bank account, and the spend you’re bleeding every month on tools nobody opens. A new deal takes a quarter to close and another to collect; an overdue invoice you’ve already earned takes a phone call, and a duplicate subscription takes a cancel. The two levers in this module move cash this week — one by collecting what you’re owed, the other by cutting what you shouldn’t be paying — and both run on evidence, not vibes.

The cash you’ve already earned

When a team needs cash, the instinct is to point at the pipeline — sell more, sell faster. But the pipeline is the slow lever: a sale is a quarter of work to close and another to collect. The fast lever is the money you’ve already earned and haven’t been paid for yet. The trouble is that it’s invisible until you lay it out — and the way you lay it out is an AR aging.

• The aging is the map. Every unpaid invoice, sorted into four buckets by how long it’s been overdue: 0–30, 31–60, 61–90, and 90+ days. The buckets aren’t bureaucracy; they’re triage. A five-day-late invoice is a reminder. A ninety-day-late one is a problem that’s quietly turning into a write-off. The aging is the single view that tells you, at a glance, how much of what you’ve earned is actually at risk — and which of it is about to stop being collectable at all.
• Rank by exposure, not just age — amount × days overdue. Chasing the oldest invoice first feels right and is wrong. The number that decides who to call is the product: the amount on the invoice times the days it’s overdue. A big invoice that’s moderately late can outrank a small one that’s ancient — a lot of money aging fast is a bigger hole than a little money aging slowly. Rank the chase list by that exposure, biggest-and-oldest first, and you spend your collections hours where the cash actually is.
• Collecting beats selling for the cash you need now. This is the mindset shift the whole module turns on. The aging is not a back-office chore; it’s the highest-yield hour in a tight month, because every figure on it is money you already booked as revenue and simply haven’t received. And it inherits the data-trust gate from M1: the aging holds customer account numbers and balances, so it stays in your approved workspace — Claude reads the local copy through the Desktop file pane, each read approved in the “Ask permissions” prompt — and every figure on it traces to the source invoice before you act on it.

The right reminder for the bucket

A reminder is the most relationship-sensitive thing finance sends, and the single most common collections mistake is sending the same one to everybody — a flat “your account is overdue” blast that treats a five-day slip from a loyal customer like a ninety-day default. The fix is a discipline a runbook would recognize — every line owns a name and a date — applied to collections: every reminder owns a real, traced invoice and a tone that fits the bucket.

• The tone escalates with the age — on purpose. A warm nudge at 0–30 (“just a heads-up, this one’s showing as open”). A clear ask at 31–60 (“can you confirm timing?”). A firm note at 61–90 (“we need payment by this date”). A firm note with a deadline and a consequence at 90+ (“payment by this date or a call to agree a plan; new orders paused”). The same balance, four registers. Send the 90-day firm note to a five-day-late good customer and you torch a relationship; send the warm nudge to a 90-day default and you waste the month.
• Every reminder cites a real invoice — traced, not guessed. This is the collections face of the VERIFY gate: you never chase a number you haven’t traced back to the actual invoice — its number, its amount, the date it was due. A reminder that says “you owe us around AED 38,000” when the real figure is 38,500 hands the customer a reason to dispute and stall. Trace it first, then chase the exact number. Claude will state an amount and a due date with total confidence and can be wrong, so the trace happens before the send, every time.
• Never a dunning blast — and a write-off is not yours to grant. A mass “PAY NOW” email to the whole ledger is the fastest way to damage relationships with the customers who were about to pay anyway. Each reminder goes to one customer, about one traced invoice, in the tone its bucket earns. And when a customer disputes and the answer is a credit note, a discount, or a write-off, that moves money — so it clears the controls gate from M1: the preparer (you, or Claude drafting) writes it; a different person named in finance-controls.md approves it. Drafting is delegated; granting money back is not.

DSO is the scoreboard

The aging tells you who to chase today. DSO tells you whether the lever is working at all. Days Sales Outstanding is the one number that turns collections from a pile of individual chases into a function you can manage — and it’s the number a board or a bank asks for when they want to know how healthy your cash conversion really is.

• DSO is how fast you turn an invoice into cash. The definition in one plain sentence: it’s the average number of days it takes to collect after a sale. The formula in one more: DSO = (accounts receivable ÷ revenue) × the number of days in the period. A DSO of 41 means a typical invoice takes 41 days to land in the bank. That’s it — no black box, which is exactly the point, because a DSO you can’t reproduce is a DSO you can’t defend to the board.
• The trend is the signal, not the number. One DSO reading is a snapshot; DSO month over month is the scoreboard. Falling DSO means the collections lever is working — you’re turning booked revenue into cash faster. Rising DSO is the early warning that AR is creeping out of control, visible months before it shows up as a cash crunch. Track it every close, right next to the aging, and you can see the lever move.
• Show the math — it’s a board number. Because DSO goes upstairs, it carries the VERIFY gate with it: state the AR figure, the revenue, the day count, and the formula, all tracing back to the close you ran in Module 2. “DSO improved” is a vibe; “DSO went 47 → 41 because we collected AED 100K of aged partner invoices this month” is a number a CFO can stand behind.

Spend you can cut with evidence, not vibes

The other lever doesn’t collect cash — it stops cash leaking. And the leaks are almost never in the big, obvious line; they hide in the gaps between tools, in the subscription nobody owns, in the renewal that auto-charged while everyone assumed someone else was watching. You can’t cut what you can’t see, so the whole move starts with putting every vendor and subscription onto one page.

• One view, every vendor. Name, category, monthly cost, seats (used versus paid), renewal date — one row each, the whole estate on a single page. The consolidation itself is most of the value: spend that’s invisible one invoice at a time becomes obvious the moment it’s all in front of you. This is the lens that turns “we spend too much on software” into a list of specific, defensible cuts.
• The three things the single view catches. Duplicate tools — two overlapping subscriptions doing the same job, where one is pure waste. Silent price hikes — the tool that crept up at renewal and nobody noticed because the charge just… continued. Unused seats — you’re paying for twelve, four belong to people who left. None of the three is visible from a single invoice; all three jump off the page the instant every vendor is on it.
• Cut with evidence, not a vibe. “We spend too much on tools” is a vibe. “ClickUp and Asana both do project management, the team standardized on Asana in March, cancel ClickUp for AED 1,200/mo” is evidence. Every cut names the tool, the reason (the overlap, the creep, or the dead seats), and the exact AED recovered — so the decision survives the meeting where someone asks “wait, who uses that?” And because a cancellation changes a contract, it’s a money move: the preparer drafts the cut list; the approver in finance-controls.md signs before anything is actually cancelled.
• A renewals calendar so nothing surprises you. Map every renewal across the year. The failure mode the calendar exists to prevent: a tool on a promotional first-year rate that auto-renews at list — sometimes 3× — because nobody calendared the date to renegotiate. A renewal you see coming 60 days out is a negotiation; a renewal that surprises you is a bill you already paid.

The Arabic collections reminders — authored, a peer lane

Here’s where most teams break the lever: they run the entire English reminder ladder, then translate it into Arabic at the end. The result is a dunning note that reads, in Arabic, exactly like what it is — a cold Western collections letter wearing a translation — landing in front of a customer whose relationship with you is personal. In the Gulf, that doesn’t just fail to collect; it can cost you the client along with the invoice.

• Author from the relationship, not from the English ladder. A collections reminder to an Arabic-reading customer is authored, warm, and relationship-first. It opens with the person and the relationship, then the invoice — never a “FINAL NOTICE — IMMEDIATE PAYMENT REQUIRED.” The reminder still does its job; it just does it the way a respected partner would raise it, because that’s the way it actually gets paid here.
• Same invoice, same trace, different register. Warm and relationship-first does not mean vague. The Arabic reminder names the exact invoice, the exact amount, the exact due date — the same VERIFY gate as the English. Numbers stay Western, currency is AED, and the register is lightly-formal MSA with Gulf-natural phrasing, per the bilingual standard you set in M1. What changes is the framing, not a single fact.
• The escalation is gentler and slower — but it still escalates. Even a 90+ Arabic note stays respectful: the firmness shows in a clear amount and a named deadline, not in capitals and threats. And the same controls gate holds across both languages — if the resolution is a credit note or a payment plan, a person other than the preparer approves it, and an Arabic note about money clears an Arabic reader who actually reads it.

Your assignment

Build the cash levers for one set of books — your own (recommended: the output is cash you actually collect and spend you actually cut) or the sample company Mizan, the GCC small-business bookkeeping SaaS whose own finance team runs throughout this track. Everything inherits the three foundation files from M1; if you don’t have them, do Module 1 first. Open the folder with your open-invoices export and your subscription list in Claude Desktop, approve each read in the “Ask permissions” prompt, and work in the chat — no terminal needed. Account numbers and AR detail stay in that workspace.

Module 4 deliverable — the cash levers

Inherit from M1: categorization-rules.md + data-checklist.md +
                 finance-controls.md (the trace-to-source rule, the
                 separation-of-duties gate, the bilingual standard)

1. ar-aging.md   (collect what you're owed)
   - every open invoice sorted into 0–30 / 31–60 / 61–90 / 90+
   - a chase list ranked by EXPOSURE = amount × days overdue
     (biggest-and-oldest first, not just oldest)
   - one reminder per chase, matched to the bucket (warm → clear →
     firm → firm+deadline), each citing a REAL invoice (number, amount,
     due date) — no untraced chase
   - failed CARD charges handled as a dunning retry, separate from the
     invoice chases

2. DSO, computed + tracked
   - DSO = (AR ÷ revenue) × days in the period — the formula SHOWN
   - this month's number AND the trend vs last month (is the lever working?)

3. vendor-spend.md   (cut what you shouldn't be paying)
   - every vendor/subscription on one page: name, cost, seats, renewal
   - the cuts, with EVIDENCE: the duplicate tools, the silent price hike,
     the dead seats — each with the exact AED/mo recovered
   - a renewals calendar for the year — so nothing auto-renews at 3× by surprise

Controls: a credit note, a write-off, or a vendor cancellation MOVES money —
drafted by the preparer, approved by someone else (finance-controls.md).

Bilingual teams: the Arabic reminders are AUTHORED, warm, relationship-first —
not the English ladder translated. Same invoice, same trace; gentler register.

The What you keep section below gives you a fill-in template for the aging and the vendor view, plus the prompts that build them — so you start from structure, not a blank ledger.

How it’s graded — the rubric

This is the part the free playbooks don’t have, and the part that makes the credential mean something. Your deliverables are scored against five criteria. Each is meets / nearly / not yet — and a “nearly” on any one is a revise, not a pass.

Cash-levers rubric — five criteria, each meets / nearly / not yet.
A "nearly" on any one is a revise, not a pass.

1. The aging is correct        Every open invoice is in the right bucket
   and ranked                  (0–30 / 31–60 / 61–90 / 90+), and the chase
                               list is ranked by exposure — amount × days
                               overdue, biggest-and-oldest first.

2. Reminders fit + trace        The reminder matches its bucket (the tone
                               escalates warm → clear → firm) and EVERY one
                               cites a real invoice — number, amount, due
                               date. No chase on an untraced number.

3. DSO is computed + tracked    Days Sales Outstanding is computed with the
                               formula shown, and reported against last
                               month so the trend is visible — not a one-off
                               number with no method.

4. The cuts have evidence       The single vendor view catches the duplicate
                               tools, the silent price hike, and the dead
                               seats — each cut named with the exact AED
                               recovered — and a renewals calendar exists so
                               nothing auto-renews at 3× by surprise.

5. The Arabic reminders are     Authored warm and relationship-first, not the
   a peer lane                  English ladder translated — same invoice and
                               same trace, gentler register, numbers Western.
                               (Bilingual teams.)

The discipline is deliberately what a senior controller would demand: an untraced chase damages a customer relationship, an unranked aging wastes the collections hour on the wrong account, and a spend cut without evidence dies in the first meeting — so the gaps a routine month absorbs quietly have to be caught here, on the page, before the reminder sends or the contract cancels.

The bar, shown — a worked model answer (Mizan)

You don’t have to guess what “meets” looks like. Here are passing excerpts for the sample company — Mizan’s own May receivables and subscriptions. Yours won’t look identical; it just needs to clear the same bar.

ar-aging.md — Mizan: open receivables (excerpt, ranked by exposure)

Customer                Invoice    Amount   Days  Bucket   Exposure
                                   (AED)    over           (amt × days)
Buraq Accounting        INV-2271   38,500    62   61–90    2,387,000
Rashid Audit Partners   INV-2188   18,500    94   90+      1,739,000
Gulf Ledger Partners    INV-2310   21,000    35   31–60      735,000
Diwan Books Co.         INV-2342   12,400     8   0–30        99,200
Najd Logistics          (card)        736     —   dunning    — (card retry)

Chase order: Buraq FIRST — biggest exposure, even though Rashid is older,
because 38,500 × 62 beats 18,500 × 94. "Oldest" loses to "oldest AND big."
Najd is a failed CARD charge, not an invoice — it goes to the dunning
retry (an update-your-card nudge), not the collections ladder.
This excerpt is the top chase rows; full-ledger AR ≈ AED 700,000.

reminder — Diwan Books Co. (0–30, warm nudge)

  Subject: Quick note on invoice INV-2342

  Hi [name] — just a friendly heads-up that invoice INV-2342 (AED 12,400,
  the Q2 subscription) was due on [date] and is still showing as open on
  our side. If it's already on its way, please ignore this. If not, a
  quick note on timing would be great — happy to resend the invoice if it
  helps. Thanks as always for working with us.

  [Traced: INV-2342, AED 12,400, due [date] — confirmed against the ledger.]

reminder — Rashid Audit Partners (90+, firm)

  Subject: Overdue invoice INV-2188 — AED 18,500, now 94 days past due

  Hi [name] — invoice INV-2188 for AED 18,500 was due on [date] and is
  now 94 days overdue. We've flagged it twice. We need either payment by
  [date] or a call this week to agree a payment plan; until it's resolved
  we'll be pausing new orders on the account. I'd much rather find a way
  through this together than escalate — can we talk this week?

  [Traced: INV-2188, AED 18,500, due [date]. A write-off or a payment plan
  is a money move → the controller approves it, not the preparer.]

DSO — Mizan, May close

  DSO = (accounts receivable ÷ revenue) × days in the period
      = (≈AED 700,000 ÷ AED 512,000) × 30  ≈  41 days

  Last month: (≈AED 800,000 ÷ AED 512,000) × 30  ≈  47 days
  Trend: 47 → 41, −6 days → the collections lever is working.
  AR, revenue, and the day count all trace to the May close (M2).

vendor-spend.md — Mizan: every tool on one page (excerpt)

The cuts (evidence, not vibes — name the tool + the exact AED)
  DUPLICATE   Asana + ClickUp both do project management. The team
              standardized on Asana in March; ClickUp is pure overlap.
              → cancel ClickUp.   Saves AED 1,200/mo  (AED 14,400/yr)
  PRICE HIKE  Amplitude (analytics) renewed at AED 2,950/mo, up from
              AED 1,950 — a silent +AED 1,000/mo nobody approved (it's the
              creep that hit May's Software line).
              → challenge / downgrade the tier.  Recovers AED 1,000/mo
  DEAD SEATS  Figma is billed for 12 seats; 4 belong to people who left.
              → drop to 8 seats.  Saves AED 300/mo  (AED 3,600/yr)
  ───────────────────────────────────────────────────────────────────
  Total identified: ~AED 2,500/mo ≈ AED 30,000/yr off the Software line —
  which is exactly the line leadership flagged for jumping this month.

Renewals calendar (next 3 — calendar each 60 days early)
  Aug 2026  Asana       annual   confirm the final seat count first
  Sep 2026  Amplitude   annual   ⚠ promo year ENDS — list price is ~3× the
                                   first-year rate; renegotiate or move NOW
  Nov 2026  Figma       annual   renew at 8 seats, not 12

Controls: a cancellation changes a contract → the preparer drafts it; the
approver in finance-controls.md signs before anything is cancelled.

ar-reminders.md — Mizan (note)

The Arabic reminders are authored, not translated. The 0–30 nudge to an
Arabic-reading customer opens with the relationship, then the invoice —
warm, never a "FINAL NOTICE." Even the 90+ note stays respectful: the
firmness is a clear amount and a named deadline, not capitals or threats.
Same invoice, same trace as the English; numbers Western, currency AED,
register lightly-formal MSA with Gulf-natural phrasing. In the Gulf a
collections note is a relationship act — a cold translation costs you the
client along with the invoice.

Notice what the model does and doesn’t do: every chase names a real, traced invoice; the chase order is exposure-weighted, not gut feel (Buraq beats the older Rashid because the money is bigger); the DSO shows its formula and its trend; every cut names the tool, the reason, and the exact AED recovered; and the Arabic reminder is authored from the relationship rather than translated from the ladder. Nothing here is a vibe — and nothing that moves money (a write-off, a cancellation) executes without the second signature.

What you keep — toolkit additions

These extend the Foundation toolkit you already installed — same folder, same CLAUDE.md, two more templates and a few more saved prompts. Drop them next to your foundation files, and every collections run and spend review from now on starts from structure instead of a blank ledger. Replace the Mizan placeholders with your own.

The ar-aging.md template

# ar-aging.md — [Company] receivables

## The buckets (every open invoice lands in exactly one)
0–30 days   |   31–60 days   |   61–90 days   |   90+ days

## The chase list (ranked by exposure = amount × days overdue)
Customer      Invoice   Amount   Days  Bucket   Exposure (amt × days)
[name]        [INV-#]   [AED]    [n]   [bucket] [product → sort high→low]
...
(Biggest-and-oldest first. Failed CARD charges go to dunning, not here.)

## The reminder ladder (match the tone to the bucket — always cite the invoice)
0–30   warm nudge   "friendly heads-up — INV-# (AED x) was due [date]"
31–60  clear        "INV-# is now [n] days past due — can you confirm timing?"
61–90  firm         "INV-# is [n] days overdue — we need payment by [date]"
90+    firm+        "INV-# is [n] days overdue — payment by [date] or a plan;
                     new orders paused" (a write-off / plan → the approver signs)

Rule: never send a reminder on a number you haven't traced to the invoice.

The vendor-spend.md template

# vendor-spend.md — [Company]: every tool on one page

## The inventory (one row per vendor — this view IS the value)
Vendor    Category   Cost/mo   Seats (used/paid)   Renews     Owner
[name]    [cat]      [AED]     [x / y]             [date]     [who]
...

## The cuts (evidence, not vibes — name the tool + the exact AED)
DUPLICATE   [tool A] + [tool B] overlap → cancel [one]    → saves [AED]/mo
PRICE HIKE  [tool] crept [+%] at renewal, unapproved      → recovers [AED]/mo
DEAD SEATS  [tool]: [n] paid seats unused                 → saves [AED]/mo

## Renewals calendar (calendar each one ~60 days early)
[month]   [tool]   [annual/mo]   [the note — flag any promo→list jump]
...
⚠ Flag any tool whose promo year ends into a 3× list price.

Controls: a cancellation / contract change MOVES money → the preparer drafts,
the approver in finance-controls.md signs BEFORE anything is cancelled.

The saved-prompt library (Cash levers) — add to the Foundation set

Age the receivables and rank the chase list
  "Read this open-invoices export [open the file] and data-checklist.md.
   Profile it first (row count, blanks, duplicate invoice IDs, date formats).
   Then build an AR aging: put every open invoice in 0–30 / 31–60 / 61–90 /
   90+ by days overdue, and rank the chase list by exposure = amount × days
   overdue, biggest-and-oldest first. Pull failed CARD charges into a
   separate dunning list. Show the days-overdue math for the top 5, and trace
   every amount to its source row — chase nothing you can't trace."

Draft the reminder for the bucket (always citing the invoice)
  "For [customer], invoice [INV-#] (AED [x], due [date], [n] days overdue,
   [bucket]): draft a collections reminder whose tone matches the bucket —
   warm for 0–30, clear for 31–60, firm for 61–90, firm-with-a-deadline for
   90+. Cite the real invoice number, amount, and due date. Keep it to one
   short message. Do NOT propose a credit note or write-off — if the customer
   disputes, flag it for the approver named in finance-controls.md."

Consolidate the vendors and find the cuts
  "Read this list of subscriptions / vendor invoices [open the file] and
   categorization-rules.md. Put every vendor on one page: name, category,
   monthly cost, seats (used vs paid), renewal date. Then flag the cuts with
   evidence — duplicate tools doing the same job, any price that rose at
   renewal, and seats we pay for but don't use — each with the exact AED/mo
   recovered. Build a renewals calendar for the year and flag any promo rate
   that renews into a much higher list price. Don't cancel anything — draft
   the cut list for a person to approve."

Author the Arabic reminder (warm, relationship-first — not translated)
  "For [customer], invoice [INV-#] (AED [x], [n] days overdue): author an
   Arabic collections reminder from scratch for a Gulf customer — warm and
   relationship-first, opening with the relationship before the invoice. Name
   the exact invoice, amount, and due date; keep numbers Western and currency
   AED. Match the firmness to the bucket but never use threats or capitals —
   even a 90+ note stays respectful. This is authored, not the English
   reminder translated."

On the Terminal & Automation track these prompts become one-keystroke slash commands; on Desktop you paste them in the chat with the right files open, approve the read in the “Ask permissions” prompt, and they do the rest — no terminal required.

What you’ve proven — and what’s next

Clear this rubric and you’ve proven the part of finance that never shows up in a single report: not closing the books, but moving the cash — collecting what you’ve already earned before it ages into a write-off, and cutting the spend you’re bleeding on tools nobody uses, every chase traced and every cut evidenced, in two languages. That’s the Cash-levers stage of “Certified Finance & Ops with Claude” — the module that turns a clean report into actual money in and waste out, which is exactly the result a team buys the certificate to vouch for.

One module remains before the capstone, and it’s the one that puts everything in front of the people who decide:

• Module 5 — the close & the board: the full month-end close run as one tight sequence, and the board pack — the figures, the narrative, and the authored-in-Arabic board story — that turns everything you’ve built into the document leadership actually decides from.

Then the capstone — Close-in-a-Box — runs all five modules end to end for a single company: one month’s books closed, reported, planned, collected-and-controlled, and put in front of the board as one connected system, graded into the credential. The two cash levers you just built are two of its moving parts — the ones that turn a tidy close into a healthier bank balance.

First, make what you built reusable. The toolkit additions above — the aging and vendor-spend templates plus the saved prompts — extend the Foundation toolkit you already installed, so your whole team runs the levers from the same files. And if you’re rolling this across a team, the operating guide is the data, controls, and sign-off layer that sits underneath all of it.