Maria Calderon

What We Observed

The window title "Initiate a Payment" in PayPal appeared 530 times across 10 working days. That is Maria manually initiating payments, one at a time, from PayPal's standard send-payment interface. This is the equivalent of sending 530 individual checks by hand instead of running a payment batch. PayPal's Payouts API exists precisely to eliminate this workflow.

The Sequence

The data shows Maria working from "PAYABLES_Schedule_2026" — her payables scheduling spreadsheet (368 events) — and executing each line item individually in PayPal. She's manually copying recipient details and amounts from the schedule into PayPal's payment form, one by one. The 420 Sheets↔PayPal context switches confirm this is not occasional — it is her primary payment workflow.

10:15 — PAYABLES_Schedule_2026 in focus (Google Sheets)
10:15 — Review next payable row
10:16 — Switch to PayPal — "Initiate a Payment"
10:16 — Type recipient email/name from schedule
10:17 — Type payment amount
10:17 — Select payment type / note
10:18 — Submit payment
10:18 — Switch back to Sheets — mark row as paid
10:18 — Move to next row — repeat

The Root Cause

Maria has a structured payables schedule that already contains all the data PayPal needs. The gap is the execution step — she's manually bridging what should be an automated batch. PayPal's Payouts API accepts a JSON payload of multiple recipients and processes them in one call. Her schedule is already the source of truth. It just needs to be wired to the API.

Implementation Path

API References

• PayPal Payouts API — send payments to multiple recipients in a single API call
• PayPal Payouts — Batch Processing — submit batch from structured JSON (maps directly from her Sheets schedule)
• Google Sheets API — read PAYABLES_Schedule_2026 rows as batch input

What We Observed

Maria's bank reconciliation workflow is centered on a Google Sheet titled "RECON as of March 12, 2026" — a date-named reconciliation file that she builds manually by cross-referencing the bank portal against Brex and other transaction sources. The pattern is a three-source loop: bank portal → Sheets → Brex → Sheets. The Ctrl+F (Find) keystroke signature is prominent during these sessions — she's searching for transaction amounts by hand.

The Structural Problem

Maria is reconciling across three sources — the bank portal, Brex (corporate card), and her own Sheets — none of which are connected. The "RECON as of March 12" naming convention also suggests she's building a new file per reconciliation period rather than using a rolling, formula-driven sheet. Each new period restarts the manual build.

Implementation Path

API References

• Brex Transactions API — pull card transactions, settlements, and cash account entries programmatically
• Brex Accounts API — account balances and statement data for reconciliation
• Google Sheets API — write Brex transaction data into the reconciliation template

What We Observed

Maria's invoice tracking workflow involves Invoiced.com as the invoice management system and a Google Sheets file titled "Invoice Monitoring(AutoRecovered)" as her working log. The "(AutoRecovered)" in the window title is a red flag — this is a file that has crashed or lost connection during active editing and been recovered by Google's auto-save. With 94 events, this isn't a one-time incident. This file is unstable and it's a core workflow surface.

The Sequence (Observed)

10:30 — Invoiced.com in focus (invoice detail)
10:30 — 10–20 second dwell (reading invoice fields)
10:31 — Switch to "Invoice Monitoring(AutoRecovered)"
10:31 — Keystroke burst: vendor name typed
10:31 — Tab key
10:31 — Keystroke burst: amount typed
10:32 — Switch to Calculator
10:32 — Tax or fee calculation
10:33 — Switch back to Sheets — total entered
10:33 — Return to Invoiced — next invoice

Why This Is Solvable Today

Invoiced.com has a full REST API and CSV export. Maria's Invoice Monitoring sheet could be populated directly from an Invoiced export — one import per day eliminates all manual re-entry. The Calculator dependency (49 opens during invoice sessions) disappears with tax formula columns in the sheet. The file stability issue resolves when the sheet becomes read-only auto-populated instead of a manual entry surface.

Quick Fix: Invoice Sheet Tax Formula

=B2*(1+D2)
// B2 = subtotal, D2 = tax rate (e.g., 0.0875)
// Or for variable rates across rows:
=SUMPRODUCT(B2:B50*(1+D2:D50))
Copy

Implementation Path

API References

• Invoiced.com Invoices API — list, filter, and retrieve invoice data programmatically
• Invoiced.com Webhooks — push invoice events to a target URL on create/update/payment
• Google Sheets API — write invoice data to the monitoring sheet

What We Observed

The window title "Daily Finance Huddle" in Google Meet appeared 60 times across the 11-day period — Maria participates in a daily finance standup. The data shows a consistent pattern: high activity in Sheets and across financial platforms in the 15–20 minutes immediately before the Meet session, indicating Maria is compiling data for the huddle in real time, not arriving with a prepared report.

What This Means

Maria is collaborative — she's in a daily finance huddle, which is a healthy operational pattern. But the value of that meeting depends on what data is available. Right now, the data is assembled during or immediately before the call. With automated reporting, Maria arrives at the Daily Finance Huddle with numbers already populated: yesterday's payments, current payables position, Brex card activity, Gusto payroll status. The meeting shifts from "let me pull that up" to "here's what I see."

The Revenue Report Signal

The window title "PARTNER MONTHLY REVENUE | March 2026" with 28 events suggests Maria is also managing partner revenue reporting. This is another manually compiled sheet that maps to a reporting cadence. It's a natural candidate for auto-population from Stripe and PayPal transaction data.

Implementation Path

API References

• Stripe Charges API — pull revenue by date range, customer, or payment method
• PayPal Transaction Search API — pull payment totals for daily summary
• Brex Transactions API — daily card spend summary for the huddle report

What We Observed

Maria has independently adopted ChatGPT — 268 events, 3.4 hours across the pilot period. This is not a finding about AI risk or shadow IT. It's a signal about a proactive, adaptive person who is already reaching for better tools. The question is what she's using it for — and how that usage could become more systematic.

What This Means

Maria is using ChatGPT the way most early adopters do — as a versatile assistant for tasks that don't have a clean workflow yet. Drafting a vendor email, figuring out a Sheets formula, looking up payroll tax rules. The instinct is right. The gap is that each session starts from scratch — she has no saved context about Summit Trail's vendors, payment terms, tax rates, or procedures. Every query rebuilds context that should already exist.

What a Structured AI Workflow Looks Like

• Vendor communication templates: Pre-built prompts for payment notices, dispute responses, and late payment follow-ups — with Summit Trail's tone and vendor names already embedded.
• Formula library: A saved set of Sheets formulas for her recurring calculations (tax, deduction verification, reconciliation) — so she retrieves instead of regenerates.
• Payroll tax assistant: A structured prompt that answers "what is the current [state] payroll tax rate for [category]" — consistent, auditable, not ad hoc.
• Exception triage: When the batch payment run or RECON sheet flags an anomaly, a structured prompt helps Maria categorize and respond consistently.

What We Observed

Maria touches every financial transaction at Summit Trail Gear. PayPal payments, Brex card charges, Gusto payroll runs, Invoiced invoices, Stripe receipts — all of it flows through her. She is the only person with a complete, daily financial picture of the business. And the data shows she's already noticing cost anomalies — informally, in the margins of her existing workflows.

The Financial Impact (P&L View)

The Time Recovery Makes This Possible

Maria already notices these patterns. The problem is she doesn't have bandwidth to act on them — she's spending 18–22 hours per week on the manual workflows documented in findings F-01 through F-06. Free that time, and the cost optimization work Maria is already doing informally becomes a structured, monthly cadence.