Fiscal-Sponsorship Platform
Fractional Technical Leadership · Solutions Architecture
The client is a fiscal sponsor — it processes expenses, donations, and payments on behalf of the sponsored projects it supports. Every project that joins brings its own vendors, bills, and reporting obligations, and the sponsor is accountable for getting all of it right. Arc Theory has led the platform's technical direction since 2023.
Growth was capped by hands-on work. Every bill was parsed out of a request's attachments into a working spreadsheet, mapped to the right vendor, class, and category, then keyed into Ramp by hand — in drop-everything batches of up to 44 bills. The team never missed its weekly payment SLA, but every new client added more manual reconciliation across Airtable, Ramp, and QuickBooks. Supporting more projects meant hiring more hands — or building a system.
Since the portal was implemented, we've grown by at least 30 projects — and we can serve all of them well because of the tools that were built.
— Katie, Operations Manager
What we built together: a single Cloud Run service that sits between Airtable, Ramp, and QuickBooks — the integration layer that automates the platform's core flow.
When a project manager logs an expense, the middleware does the rest: it creates the bill, tags it to the correct organization, class, and vendor, then creates and issues the payment — keeping Airtable, Ramp, and QuickBooks reconciled across the entire flow. The manual cross-platform work that used to cap capacity simply disappears.
- Create vendorRamp
- Issue paymentRamp
- Create classQuickBooks
- Account-balance reportQuickBooks
The platform went live in April 2025. The numbers below compare the manual era to the platform era — same team, same weekly SLA.
Yearly transactions climbed 59% — from 1,857 to 2,959. Dollars processed rose alongside them, up 46% from $5.94M to $8.66M.
The platform now serves an average of 43 active clients a month, up from 22 — nearly double, with the same team and the same weekly SLA.
Average vendors paid each month rose from ~45 to ~64 after the platform went live — same team.
Same Friday SLA — a fraction of the human time.
Payments always went out on time — the team never missed its weekly SLA. What the data shows is what hitting it used to cost. Under the legacy process every bill was parsed out of a request's attachments into a working spreadsheet, mapped to the right vendor, expense class, and category, then keyed into Ramp by hand — in drop-everything batches of up to 44 bills. The platform creates the bills itself; a human now reviews the expense and approves.
11.8 minutes of parsing, mapping, and keying → 2 minutes of review
Same team, same Friday SLA — ~34 bills/week in 2024 (fully manual), ~45 through the 2025 stabilization year after the April 2025 launch, ~52 in 2026 now that the platform has fully taken over. Hands-on time fell from roughly 6½ hours a week to under 2.
- Triage request, download attachments2.8 min
- Parse attachment into spreadsheet3.0 min
- Match vendor (or set up new)2.0 min
- Map expense class + category1.5 min
- Key bill into Ramp, attach receipt2.0 min
- Schedule payment & verify0.5 min
- Review expense & approve2.0 min
Bill creation, vendor matching, class & category coding, payment scheduling, and the Ramp + QuickBooks sync all happen automatically.
Source: process-optimization analysis of 937 legacy payout requests (~2,641 manually keyed bills, Mar 2023–Apr 2025) and 2,832 platform-created bills (piloted Feb 2025, launched Apr 7, 2025), reconciled against Ramp records; legacy entry timing validated against Ramp API timestamps. Weekly throughput counts actual Ramp-recorded bills at steady state, not an attachment extrapolation. Per-bill minutes are a task-level model — Ramp keying measured from timestamps, remaining steps estimated and corroborated by AP industry benchmarks.
H1 2026 is already up 40% on H1 2025 — and the platform keeps adding volume.
The platform now supports nearly twice the clients each month — an average of 22 before the system, 43 after — with no added headcount. The manual reconciliation across Airtable, Ramp, and QuickBooks that capped that number is automated end to end.
Source: platform reconciliation, names and PII removed. Full-year figures cover complete years; H1 (Jan–Jun) is used to compare the in-progress 2026 fairly, and still undercounts it.