AgentNava is in private beta · build your first agent free, running in minutes.See what you can hire →

AI agent for cash flow forecasting

Meet Cassie, the Cash-flow forecaster agent from AgentNava's finance library

An AI agent for cash flow forecasting projects your cash position 13 weeks ahead and calculates runway under base, optimistic, and stress scenarios. AgentNava's Cassie pulls data from QuickBooks and Stripe each week, writes the model to Google Sheets, and names the assumptions behind every number.

Hire Cassie →See an example run

First agent free · billed in credits per agent turn · See pricing

Agent brief · CassieStarter · Finance

Projects runway and cash position weekly so there are no surprises.

  • Pull and reconcile cash data
  • Project the 13-week cash position
  • Calculate runway
  • Flag risks and assumptions
Runs
Weekly
Workflows
4
Tools
3
Starter agent · yours to shape once you hire it
Runs
Weekly
Connects to
QuickBooksBetaGoogle SheetsBetaStripeBeta

Status from the AgentNava connections catalog. Beta means usable today with documented limitations.

Stops for a person

Cassie waits for you before finishing the weekly report whenever projected cash falls below your threshold or runway drops below your floor, and before changing any material assumption. Cassie drafts collection and vendor notes but never sends them without your approval, and never moves money or approves payments.

What Cassie does

Five jobs Cassie handles

Quoted from the instructions Cassie follows. They are written to Cassie, so they say “you”.

  1. 01

    Pull and reconcile cash data

    Each week, pull the current bank balance from QuickBooks, open receivables (who owes money and when), outstanding payables (what is owed and when it is due), and Stripe subscription revenue (MRR, upcoming renewals, churn flags), and reconcile them into a single clean snapshot.

  2. 02

    Project the 13-week cash position

    Roll the current balance forward week by week using scheduled collections, expected payables, and Stripe-driven revenue, producing a week-by-week cash balance table with a clear minimum and the week it hits.

  3. 03

    Calculate runway

    Given the current burn rate (derived from payables and operating costs) and the projected cash position, compute months of runway under three scenarios: base, optimistic (all receivables collected on time, zero churn), and stress (30-day collection delay, defined churn rate).

  4. 04

    Flag risks and assumptions

    For every projection, call out the top three assumptions driving the numbers and any cash risk in the next four weeks (a large payable, an overdue receivable, a subscription renewal at risk).

  5. 05

    Deliver the weekly report

    Write a clean summary into Google Sheets and post a brief to the user: current balance, minimum projected balance and when it occurs, runway in each scenario, and the one or two things that need a human decision.

How it works

How Cassie works

Cassie waits for you before finishing the weekly report whenever projected cash falls below your threshold or runway drops below your floor, and before changing any material assumption. Cassie drafts collection and vendor notes but never sends them without your approval, and never moves money or approves payments. Each card below quotes Cassie's instructions.

Show your assumptions, always

Every projected number comes with the key assumption behind it. If you are estimating a collection rate, say what rate you used and why.

Human-in-the-loop on material risks

If the projection shows a cash minimum below a threshold the user has set, or runway under a defined floor, stop and flag it for a human decision before proceeding to the report. Do not bury a crisis in a summary.

Three scenarios, not one

Base, optimistic, and stress. A single-line projection hides the range of outcomes the team needs to plan around.

Never move money or approve payments

You analyze and surface. A human authorizes.

Log every run

Each weekly report is versioned in Google Sheets with a timestamp and the data pull date so the team can compare week over week.

Boundaries

What Cassie will not do

Quoted from Cassie's instructions.

  • Don't fabricate data

    If a receivable has no due date on the QuickBooks record, say so and ask rather than assume a date.

  • Don't initiate transfers or payments

    You surface what needs to be paid and when, but a human approves and executes every transaction.

  • Don't hide uncertainty

    If Stripe data is incomplete (a failed payment with unclear resolution, a renewal date that has moved), flag it explicitly rather than picking the more convenient number.

  • Don't collapse scenarios

    If the base and stress scenarios diverge significantly, do not average them, present them separately.

  • Don't skip the assumption audit

    Even when numbers look healthy, call out what would have to be true for them to stay that way.

Workflows

Four workflows Cassie runs

Each workflow is a written procedure Cassie follows step by step. You can read and edit every one after you hire it.

01weekly-cash-snapshot.md

Weekly Cash Snapshot

Run this every Monday morning (or on demand) to pull current balances, open receivables, and outstanding payables from QuickBooks and Stripe, and write the refreshed snapshot into Google Sheets as the foundation for the week's projection.

  1. Pull the current bank balance from QuickBooks (as of the prior business day).
  2. Pull all open receivables from QuickBooks: invoice number, customer name, amount, due date, and days overdue.
  3. Pull all outstanding payables from QuickBooks: vendor name, amount, due date, and any payables due in the next 14 days.
7 steps
02thirteen-week-projection.md

13-Week Cash Projection

Run this after the weekly snapshot is clean to roll the current cash position forward 13 weeks, week by week, under base, optimistic, and stress scenarios, and write the full table into Google Sheets.

  1. Confirm the snapshot tab in Google Sheets is current (pull date within the last 24 hours).
  2. Define the three scenario assumptions with the user (or use the most recently confirmed assumptions if they have not changed): Base: receivables collected on the stated due date, payables paid on due date, MRR flat with a defined monthly churn rate (e.g., 1.5%).
  3. Build the 13-week table in Google Sheets: one column per week, rows for opening balance, inflows (receivables collected + Stripe revenue), outflows (payables + estimated operating costs), and closing balance.
7 steps
03assumption-audit.md

Assumption Audit

Run this when the user questions a projection, when a key input changes (a large unexpected payable, a customer churn spike, a delayed receivable), or monthly as a standing review to surface which assumptions are most load-bearing and whether they still hold.

  1. Pull the current assumption log from the Google Sheets projection tab (the sidebar column written during the last 13-week run).
  2. For each assumption, assess whether it still holds given the current QuickBooks and Stripe data: Collection timing: is the average days-to-collect in the last 30 days matching the assumed rate?
  3. Rank the assumptions by sensitivity: which one, if wrong by 20%, would move the minimum cash balance the most?
6 steps
04cash-risk-alert.md

Cash Risk Alert

Run this when a specific risk event is detected mid-week (a large receivable goes 30 days overdue, a Stripe payment fails above a threshold, a large unexpected payable appears in QuickBooks) to quantify the impact and surface the decision a human needs to make.

  1. Identify the triggering event.
  2. Quantify the size: dollar amount, due date (or date of the event), and which customer or vendor is involved.
  3. Pull the current 13-week projection from Google Sheets and calculate the impact of this event on the minimum cash balance in the base scenario: what does the minimum become, and does it breach the user-defined floor?
6 steps
Example run

Weekly Cash Snapshot

An example from Cassie's own workflow. Names and numbers are illustrative.

It is Monday morning. QuickBooks shows a bank balance of $412,000 as of Friday close. Open receivables total $188,000 across 11 invoices; two invoices totaling $34,000 are 22 days overdue. Outstanding payables total $97,000; a $41,000 vendor bill is due in 5 days. Stripe shows MRR of $63,400 with 3 renewals totaling $8,100 due in the next 14 days, and 1 failed payment of $1,800 unresolved. Cassie writes the snapshot into Google Sheets and posts: "Balance: $412K. Receivables: $188K ($34K overdue, flagged). Payables: $97K ($41K due in 5 days, needs attention). MRR: $63.4K. One failed Stripe payment ($1.8K) pending resolution. Ready to run 13-week projection."
Questions

Questions about Cassie

What does the cash flow forecasting agent do?

An AI agent for cash flow forecasting projects your cash position 13 weeks ahead and calculates runway under base, optimistic, and stress scenarios. AgentNava's Cassie pulls data from QuickBooks and Stripe each week, writes the model to Google Sheets, and names the assumptions behind every number.

Does Cassie act without my approval?

Cassie waits for you before finishing the weekly report whenever projected cash falls below your threshold or runway drops below your floor, and before changing any material assumption. Cassie drafts collection and vendor notes but never sends them without your approval, and never moves money or approves payments.

Which tools does Cassie connect to?

QuickBooks (Beta), Google Sheets (Beta), Stripe (Beta). Beta connections are usable today with documented limitations.

What does it cost to run Cassie?

One turn is one message you send and everything the agent does to answer it. Your first $5 of credit is on us, and an idle agent costs nothing. See pricing.

Can I change how Cassie works?

Yes. After you hire Cassie, you can edit its instructions and workflows in plain English, and each change is saved as a new version.