All case studies
Case Study SMSF Admin & Accounting

Arithma

From three hours to twenty-five minutes per new fund.

A four-partner SMSF-focused practice in Sydney with an accounting arm. Around 320 funds on the book, 11 staff, and a principal who was tired of spending Saturdays on onboarding paperwork. We ran the Audit, shipped two sprints in eight weeks, and moved them onto a monthly retainer.

arithma.com.au * Illustrative case study. Arithma is a representative engagement.
arithma.internal / onboarding / fund-2847
Member detected
2 of 2
Confidence
94%
Ready for review
−88%
New-fund onboarding time
~620
Staff hours reclaimed / year
8
Weeks start to retainer
1.4x
Funds handled per admin

The challenge

Arithma\u2019s growth was bottlenecked by two of its most repeatable workflows. Opening a new SMSF took three hours of admin time. Month-end broker reconciliation burned 15 hours across two staff every single week.

The principal had already watched a Big-4-adjacent consultancy pitch him a six-month, A$80k "AI strategy" engagement. He passed. He wanted a working tool, not a deck.

Our approach

Started with a two-week Audit: shadowed five staff, ranked opportunities, priced them. Two clear wins jumped out \u2014 both with ROI under a year.

Shipped them as fixed-scope sprints. Integrated with the tools they already run (Class Super, Xero, SuiteFiles). No rip-and-replace, no SaaS upsell, no retainer pressure until the first sprint was live and accepted.

Timeline

Intro call to live retainer in nine weeks

Week 0

Intro call

The family-friend introduction. One 40-minute Zoom. We agreed on the Audit scope and pricing on the call.

Weeks 1–2

Opportunity Audit

Half-day onsite in their office. Five staff shadow sessions. Stack review across Class Super, Xero, Outlook, SuiteFiles. Written report ranked 5 opportunities by ROI, effort, and risk.

Weeks 3–6

Sprint 1 — SMSF Onboarding Assistant

Built an intake tool that extracts member, trustee, and contribution data from messy new-fund packs (PDF + scanned paper) into Class-ready structured data. 94% field accuracy on first pass; 100% after staff review.

Weeks 7–8

Sprint 2 — Broker Statement Reconciliation

AI pre-codes broker statement transactions against the chart of accounts. Flags exceptions for human review. Cut weekly reconciliation cycle from two full days to about six hours.

Week 9+

Ongoing retainer

Monthly tuning, evals, and two new workflows in the backlog (audit pack prep, contribution-cap correspondence). Standing office-hours slot every second Tuesday.

What we shipped

Two sprints. Two measurable wins.

SMSF Onboarding Assistant

Problem

Opening a new fund took ~3 hours: reading a trust deed, re-keying member/trustee details, chasing signatures, creating the Class file.

Solution

Tool ingests the new-fund pack (PDF, scanned, mixed), extracts structured data for Class, and drafts the missing-information checklist email automatically. Staff reviews, corrects, approves.

Outcome
  • Onboarding time: 3 hrs → 25 min per fund
  • Typos/re-keys down ≈ 95%
  • Staff can onboard while on a call, not after hours
Stack
Class Super API Claude (Sonnet) Google Document AI SuiteFiles

Broker Statement Reconciliation

Problem

Two admins spent 15 hours a week pre-coding broker statements line-by-line against the chart of accounts before month-end.

Solution

Tool parses broker CSVs and PDFs, matches to the chart using the firm’s own historical coding, flags anything under 90% confidence for human review. Posts back to Xero.

Outcome
  • Cycle time: 15 hrs → 6 hrs / week
  • Exception rate stable at ~7%
  • Month-end pressure noticeably reduced
Stack
Xero API Claude (Haiku + Sonnet) Postgres Custom eval harness
“I had an "AI strategy" deck from a bigger firm quoting A$80k and six months. Westringia ran the Audit in two weeks and had a real tool running in our office eight weeks later. My staff stopped dreading Mondays.”
Principal, Arithma

Want the same nine-week arc for your practice?

Start with an Audit. You\u2019ll get a ranked list of the two or three things worth doing, whether or not we end up building them.

Book an Audit \u2014 A$4,500