The Shift Log
Working notes, in public.
What AI does well in a small business, what it doesn't, and where it goes wrong. Written for the owners and principals who have to sign off on it. Published Mondays and Thursdays.
Workflows
15 notes
- Document assembly without a six-figure systemEngagement letters, quotes and contracts rebuilt from last month's file. How templates and merge fields fix it, and where AI helps or adds risk.
- Payday Super: the four things it breaks in your payrollSuper is now due within seven business days of every payday. What that breaks in a process built around quarterly, and how to work out your own margin.
- Stopping no-shows: booking reminders for gyms and clinicsEmpty appointment slots cannot be resold after the fact. How to price your no-show rate, what reminder cadence actually works, and what to keep manual.
- Quoting on the spot: automating quotes for trade jobsQuoting is two jobs: writing it and chasing it. What a trade or service business can safely automate in each, and the arithmetic that says whether to bother.
- Five workflows trade and retail businesses leak time onThe five places hours leak out of trade, retail, hospitality and fitness businesses, what each one costs, and which are worth automating first.
- AI meeting notes: what to automate, what to keepHow professional firms record, summarise and file meeting notes with AI: the four stages, the usual failure points, and the parts that must stay human.
- What happens when an automation breaks?Every automation eventually fails. Fail-safe design decides whether you find out in ten minutes or ten weeks. What monitoring, alerts and support cover.
- Quote accepted, invoice raised, payment chasedHow agencies and consultancies close the cashflow loop: an accepted quote becomes an invoice and a scheduled follow-up, with Xero or MYOB behind it.
- Status updates without chasing the teamHow to pull project and matter status from the systems your team already uses, instead of asking people for updates that arrive late and slightly wrong.
- AI inbox triage: classify, route, draftHow to set up AI triage on a shared or reception inbox: sensible categories, deterministic routing rules, drafted replies, and what to never auto-send.
- The five workflows most firms leak time onEnquiry intake, quote to invoice, review requests, owner reporting and inbox triage: where each leaks hours in a firm, and what good looks like.
- Building an owner's report that assembles itselfHow to build a Friday owner's report that pulls itself together from timesheets, CRM, inbox and accounting, and why a person still reads it first.
- Automating review requests without risking your reputationHow to trigger review requests when a job or matter closes, and route unhappy replies to a person before anything unpleasant lands on a public page.
- Client onboarding for accounting firms, automatedHow to automate accounting client onboarding end to end: engagement letter, ID verification, software access, document requests and first-meeting prep.
- A 30-day AI plan for a professional firmA week-by-week plan: audit where the hours go, pick one workflow, build it properly, and measure whether it worked. Written for small professional practices.
Risk & Compliance
10 notes
- Records that survive a wage claimWhat Fair Work requires employers to record and keep for 7 years, why weak records shift the burden of proof, and where automation helps or harms.
- You have to disclose the robot from 10 DecemberFrom 10 December 2026, privacy policies must say where software makes or heavily shapes decisions about people. How to find yours and what to write.
- The privacy policy that survives an OAIC sweepWhat APP 1 requires a privacy policy to say, why the OAIC's first sweep matters beyond its six sectors, and how website forms change what yours must cover.
- The AML change that quietly ended your privacy exemptionBecoming a Tranche 2 reporting entity removed the small business privacy exemption for your AML/CTF work, whatever your turnover. What that means in practice.
- AML/CTF Tranche 2: what a small firm actually has to doTranche 2 obligations started on 1 July 2026 and the enrolment deadline was 29 July. What to do if you missed it, and what a system can assemble for you.
- What an automation can and cannot do in a conflict checkA conflict check is a search problem wrapped around a judgement call. Automation belongs in one of them. Where the line sits and how to build the search half.
- Who actually owns your automations?Accounts, API keys, workflow files and code: what your firm needs in its own name, the lock-in patterns to watch for, and what to ask any provider.
- Is it safe to use AI in a law firm?Whether AI is safe in a law firm depends on what data goes where, whose account it lands in, and who checks the output before a client ever sees it.
- How to automate client intake in a law firmA practical intake workflow for a small law firm: structured form, conflict check prompt, matter record and a drafted acknowledgement that waits for approval.
- AI and client confidentiality: the practical rulesA working checklist for professional firms: de-identification, the vendor questions on residency and retention, subprocessors, and engagement terms.
Bottlenecks
9 notes
- The first 90 seconds after an enquiry landsWhy the reply to a new enquiry matters more in its first minutes than its wording, how to price slow replies, and what a fast, human-approved response needs.
- Rostering across sites without a phone treeMulti-site rosters keep routing through the owner's phone. What the daily reshuffle costs, which parts a system can take over, and what to hand across first.
- AI receptionist or answering service: what trades needThree ways to stop losing calls, and the honest case for each. What an AI receptionist does well, where a human service still wins, and when voicemail is fine.
- What a missed call actually costs a trade businessMost trades know they miss calls and almost none know what it costs. The four numbers to count, the arithmetic, and what to change once you have it.
- What an AI consultant charges in Sydney (2026)How AI consultants in Sydney price their work in 2026: hourly, day rate, fixed project or retainer, the published numbers, and how to compare quotes.
- What AI automation actually costs in Australia (2026)The five cost layers behind an AI automation: build, platform fees, model usage, maintenance and internal time, plus how to calculate the payback yourself.
- Why the partner becomes the bottleneck in a small firmHow to tell when the principal has become the constraint in a professional firm, what that costs in capacity and risk, and which habit to change first.
- Automate, buy software, or hire someone?A decision framework for small firms: when automation is genuinely right, when off-the-shelf software wins, when to hire, and when to do nothing yet.
- What admin actually costs a professional firmA method for measuring the hours your firm loses to non-billable admin, choosing an honest hourly rate, and testing whether fixing it is worth the money.
Adoption
4 notes
- Your automation is messaging staff at 9pmThe right to disconnect covers automated texts and alerts, not just a manager's call. How to set send windows, routing and quiet hours before it's a complaint.
- An owner's Monday, hour by hourOne owner's Monday morning, hour by hour, before and after the routine reading and drafting is handed to a system, and how to price your own Monday.
- Two-thirds are using AI. Five per cent get anything backMost Australian small firms now use AI and almost none have built anything around it. What separates the five per cent, and how to test one task.
- Getting your team to actually use a new workflowMost automations fail on people, not technology. How to design defaults, run training that works, close the old path, and measure whether the new way stuck.
Myths
3 notes
- Not for a business like mine? AI in trades, clinics and shopsAI won't lay tiles or treat a patient, but trades, clinics and shops carry plenty of admin it fits. Three ordinary Tuesdays, and what genuinely doesn't fit.
- It sounds right and it's wrong: where AI mistakes cost youAI writes wrong answers in the same confident tone as right ones. Where that costs a small firm money, and a review rule that doesn't eat the time saved.
- What AI won't do for your business, and what it willSix common beliefs about AI, from 'it will replace my staff' to 'it's not for a business like mine', corrected plainly, with what it does well.
Tools
2 notes
- Which AI model should your firm actually use?Claude, ChatGPT, Gemini or an open-weight model: how a professional firm should decide, based on data handling, account ownership and cost control.
- n8n, Make or Zapier: which fits a professional firm?Choosing between n8n, Make and Zapier for a professional firm: hosting and data residency, the shape of the billing, complexity ceiling, and who maintains it.