# Shift — https://shiftconsulting.io

> Shift installs practical AI workflows and websites for agencies, sole operators and
> small white-collar service businesses in Australia. Sole operator, Sydney, two active
> builds at a time. Fixed scope, built in the client's own accounts, handed over in the
> client's name. Regulated practices (solicitors, tax agents) are the sharpest case for
> the Verify step, but the work is not limited to them.

## Pages

- [Home](https://shiftconsulting.io/): the leak arithmetic, and a calculator for what admin is costing a business.
- [What we build](https://shiftconsulting.io/services/): Verify, Automations, Websites and Care, with prices.
- [How we work](https://shiftconsulting.io/how-we-work/): a worked example end to end, the six operating rules, what the client owns.
- [About](https://shiftconsulting.io/about/): who Shift is and why the intake is capped.
- [Book a fit call](https://shiftconsulting.io/contact/): thirty minutes, no deck.
- [The Shift Log](https://shiftconsulting.io/log/): working notes, published Mondays and Thursdays.

## The engagement ladder

- **⇧V Verify** (from A$5,000, two weeks) — tool inventory, platform and contract verification, data-flow map, tool register, and a written answer for clients, insurers and tender panels. Documents the facts; does not interpret obligations and is not legal advice.
- **⇧A Automations** (A$4,500–14,000, 2–5 weeks) — Flow Map (from A$750, credited against a build within 30 days), Single Flow (A$2,400–4,500), Flow System of three to five connected workflows (A$6,500–12,000).
- **⇧W Websites** (from A$4,500) — custom, mobile-first, forms wired into the workflows.
- **⇧C Care** (from A$1,500/month, optional, cancellable monthly) — monitoring, one change a month, register kept current, annual re-verification.

## Operating rules

- The client owns all accounts, keys, workflows and code from day one.
- Scope is fixed in writing before any build.
- A human reviews anything that touches a client or money.
- A broken workflow stops and alerts; it never guesses or silently retries.

## Articles (21)

- [A 30-day AI plan for a professional firm](https://shiftconsulting.io/log/30-day-ai-plan-professional-firm/): Spend week one measuring where admin time actually goes, week two choosing a single workflow and mapping it, week three building it in your own accounts with a human review step, and week four running it against your baseline and deciding whether to keep it. One workflow finished beats five started.
- [What admin actually costs a professional firm](https://shiftconsulting.io/log/admin-hours-cost/): Measure the hours first with a two-week log, then apply a rate you can defend: the fully loaded cost of the person doing the work, or the charge-out rate only for hours you could genuinely convert to billable work. Multiply hours by working weeks by rate, then discount it by the share of freed time you would realistically recover. The result is a range, not a figure, and that is fine.
- [AI and client confidentiality: the practical rules](https://shiftconsulting.io/log/ai-client-confidentiality/): Decide what information may leave the firm before anyone chooses a tool, de-identify by default, and only use identified client material inside a system the firm has a written agreement for. Then ask the vendor where data is processed, how long it is kept and who else touches it, and keep the answers on file.
- [AI meeting notes: what to automate, what to keep](https://shiftconsulting.io/log/ai-meeting-notes-professional-firms/): Yes for the mechanical parts, no for the decisions. Recording, transcribing, drafting a summary and filing it to the right client are all machine work, and they are where the hours go. Deciding whether to record, whether the summary is accurate, and what belongs on the file are professional calls that stay with the person who attended the meeting. The consent question comes first, because recording rules in Australia differ by state and by who is on the call.
- [How to automate client intake in a law firm](https://shiftconsulting.io/log/automate-client-intake-law-firm/): Yes, if the automation handles the movement of information and a lawyer keeps every decision. A workable build is: structured enquiry form, matter or prospect record created automatically, a conflict check assembled and put in front of a lawyer to decide, and a drafted acknowledgement that waits for approval before it is sent. Prospective client information is confidential from the moment it arrives, so build it in your own accounts with access limited to the people who need it.
- [Client onboarding for accounting firms, automated](https://shiftconsulting.io/log/automate-client-onboarding-accountants/): Chain five steps so each one triggers the next: engagement letter out and signed, identity and verification collected through a secure portal, software and authority access arranged, the document request issued with chasing built in, and a first-meeting brief assembled from what came back. Keep the risk assessment and the scope decision with a person, and handle identity documents as the most sensitive data in the practice.
- [Automating review requests without risking your reputation](https://shiftconsulting.io/log/automate-review-requests/): Fire the request from a real completion event, suppress any client with an open complaint, dispute or unpaid invoice, and give every client a private reply path that alerts a person immediately. The automation handles timing and sending; a human handles anything that sounds unhappy, and handles it the same day.
- [Building an owner's report that assembles itself](https://shiftconsulting.io/log/automated-owner-report/): Pick eight to twelve numbers that would actually change a decision, identify one authoritative source for each, and have a scheduled job pull them into a fixed template every week. Keep the numbers computed rather than written by a model, and have a person read the first month of reports against the source systems before anyone trusts them.
- [Automate, buy software, or hire someone?](https://shiftconsulting.io/log/build-vs-buy-vs-hire/): Decide on three things: how often the work happens, how stable the process is, and how much judgement it needs. High volume, stable and low judgement means automate; a common problem a vendor already solves means buy; work needing judgement, relationships or accountability means hire. If the process is undocumented or changes every few months, none of the three is right yet.
- [The five workflows most firms leak time on](https://shiftconsulting.io/log/five-workflows-firms-leak-time/): Almost every firm leaks time in the same five places: enquiry intake, quote to invoice, review requests, the owner's report and shared inbox triage. Start with whichever one is costing you the most hours or losing you the most work, which is usually intake or invoicing. Fix the routing and the assembly first and leave the judgement calls with a person.
- [Getting your team to actually use a new workflow](https://shiftconsulting.io/log/getting-your-team-to-use-it/): Make the new way easier than the old way, then close the old way. Adoption fails when the automation is optional, when the person who does the task was not involved in designing it, and when nobody says what happens to the time it saves.
- [AI inbox triage: classify, route, draft](https://shiftconsulting.io/log/inbox-triage-ai/): Classify each message into a small fixed set of categories, route it using deterministic rules wherever a rule exists and the classifier only for what is left, and have replies written into the drafts folder for a person to approve. Nothing involving advice, money, deadlines or complaints should ever send itself.
- [Is it safe to use AI in a law firm?](https://shiftconsulting.io/log/is-ai-safe-for-law-firms/): It can be, but safety is a property of the arrangement rather than the tool. What matters is which information leaves the firm, whose account it lands in, what the provider is permitted to do with it, and whether a person reviews the output before it reaches a client or a court.
- [n8n, Make or Zapier: which fits a professional firm?](https://shiftconsulting.io/log/n8n-vs-make-vs-zapier/): Choose on operating model, not features. Zapier suits a handful of simple connections owned by a non-technical person, Make suits moderate complexity in the cloud, and n8n suits firms that want self-hosting, control over where data sits, or enough volume that per-task billing stops making sense. All three will connect the tools a typical firm already uses, so the deciding factor is usually maintenance and hosting rather than capability.
- [Why the partner becomes the bottleneck in a small firm](https://shiftconsulting.io/log/partner-bottleneck/): You are the bottleneck when work stops moving while it waits for you. Quotes, approvals, replies and reports all queue behind one person's attention, and the firm's capacity is capped at whatever that person can personally get to in a week. The test is not how busy you feel, it is how much work in the firm cannot progress without you.
- [Status updates without chasing the team](https://shiftconsulting.io/log/project-status-updates-without-chasing/): Derive status from signals your systems already produce as a by-product of the work, such as task movement, file activity, calendar entries and invoicing, then report exceptions rather than everything. Ask a person for one judgement field per project per week, and treat a project with no signal at all as unknown rather than fine.
- [Quote accepted, invoice raised, payment chased](https://shiftconsulting.io/log/quote-to-invoice-automation/): Make acceptance an event a system can see, have that event create the job and a draft invoice in your accounting file, then schedule the reminder sequence at the moment the invoice is sent rather than when someone notices it is overdue. Payment marks the whole sequence complete automatically. Keep a human approval on the invoice until the drafts have been right for a month.
- [What AI automation actually costs a small firm](https://shiftconsulting.io/log/what-ai-automation-costs/): There are five costs, not one: the one-off build, the automation platform subscription, metered model usage, ongoing maintenance, and your own team's time during and after the build. The build is usually the largest single number but the smallest long-run one. Work out what the process costs you today in hours, price all five layers against that, and only proceed if the payback lands inside about twelve months.
- [What happens when an automation breaks?](https://shiftconsulting.io/log/when-automation-breaks/): A well-built workflow stops and alerts a named person. It does not guess, retry silently or half-finish a job. The failures that hurt a firm are the quiet ones, where the workflow keeps running and corrupts data for weeks before anyone notices.
- [Which AI model should your firm actually use?](https://shiftconsulting.io/log/which-ai-model-for-your-firm/): For most professional firms the choice matters far less than the account you use it through. Pick a business or API tier rather than a consumer plan, hold the account and keys in the firm's name, and standardise one model per workflow so the output stays consistent. Then test two providers on your own real work for a fortnight and let that decide, because capability differences shift with every release while data handling and cost control do not.
- [Who actually owns your automations?](https://shiftconsulting.io/log/who-owns-your-automations/): Only if the accounts, API keys, workflow files and code are in your firm's name, and you have documentation good enough for someone else to maintain them. Plenty of arrangements look like ownership and are actually a rental, which you discover on the day the relationship ends. Ask whose accounts the build runs in before the work starts, because that question is almost impossible to fix afterwards.

## Contact

sam.k@shiftconsulting.io