Writing a Treasury Policy Once Reserves Cross Eight Figures

Write the policy around three tiers of cash rather than a list of products: operating cash you'll need in the next 30 to 60 days, a reserve buffer you want liquid but not touched daily, and strategic reserve you won't need for six months or more. Only after you've sized those three tiers should you assign instruments to each one, because the instrument list changes as rates move but the tiers don't.

Sizing the operating tier usually starts from your burn: a company with a high burn multiple needs a deeper operating buffer than one that's closer to funding itself out of revenue, since a bad month costs it more runway¹. Say your monthly operating spend is $600,000; a company at the riskier end of that burn spectrum should hold a wider multiple of that figure liquid than one closer to breakeven.

The policy should cap how much sits at any single bank, insured or not, and how much sits in any single instrument type. A common approach: no more than a stated share of total reserves at one banking relationship, and a separate cap on any instrument that isn't a direct Treasury obligation or an insured deposit. Write the caps as percentages of total reserves so the policy doesn't need rewriting every time the balance changes.

Don't leave this section vague. "Diversify appropriately" isn't a control; a specific percentage per counterparty is, and it's the line an auditor will actually test against your bank statements. Instrument caps matter just as much: a policy that allows unlimited allocation to, say, corporate commercial paper alongside Treasury bills is really two different risk profiles hiding under one label.

Spell out dollar thresholds for single-signer approval, dual-signer approval, and board notification, and tie them to your bank's actual permission settings, not just to a paragraph in the policy. If the bank still lets one person wire a large sum alone because nobody configured the dual-approval rule, the written policy doesn't protect you.

Include a named backup approver for every threshold. A policy that only works when one specific person is reachable isn't a control, it's a bottleneck, and it tends to fail on exactly the day someone is unreachable and a payment is time sensitive.

Write these approval rules into the policy and the bank settings:

• A dollar threshold for single-signer approval, matched to the permission settings the bank actually enforces rather than a paragraph alone.
• A higher threshold that requires dual-signer approval, configured in the bank so one person cannot wire a large sum alone.
• A level above which the board or audit committee is notified before or after the money moves.
• A named backup approver, so the approval process still works when the usual signer is unreachable.

State how often the board or audit committee reviews actual holdings against the policy, typically quarterly, and what triggers an off-cycle review, such as a bank downgrade or a material change in cash balance. A policy without a stated review cadence tends to get written once and forgotten until an auditor asks for it.

Attach the review to an existing board meeting rather than inventing a new one, or it quietly stops happening after the second quarter. A one-page holdings summary that maps directly to the policy's tiers makes that review take minutes instead of becoming its own agenda item.

The most frequent gap between a written policy and actual practice isn't a rule violation, it's a policy that never got updated after a fundraise doubled the balance overnight. Say a concentration limit was originally written around a $10M balance: that same percentage cap can become meaningless once the balance triples to $30M, if nobody goes back and revisits the dollar figures behind it. Build a trigger into the policy itself instead: any raise above a stated size forces a policy review within a set number of days, rather than waiting for the next scheduled quarterly check.

The finished policy should let anyone on the finance team answer, without checking three logins, how much cash the company holds, where it sits, and whether that's within the approved limits. If your controller has to open five bank dashboards to answer that question, the policy exists on paper but isn't actually governing anything yet.

Many finance teams route the monthly reconciliation of actual holdings against the policy through an AI assistant like Frank, MeetMyCFO's AI CFO, so the variance gets flagged before the quarterly board packet gets built rather than during it.