AI Readiness Review
A scorecard and a roadmap you keep whether or not we go further.
from $1,800 + GST
1 week · fixed price
View on pricing →Human-Led AI
Your people in charge. The boring work handled.
Most businesses aren't sceptical of AI. They're sceptical of handing a process to a machine and hoping nothing goes wrong. Fair. So don't. Human-Led AI means the AI does the work, a named person owns the outcome, and the AI earns more room only as it proves itself on your real work.
Built on Microsoft 365 you already own · Every artefact belongs to you · Fixed prices, published
Three rules. Every assistant I build follows them.
A person owns every outcome.
Nothing goes to a client, a supplier or your accounts without a name attached to it. The AI prepares; a person is accountable. That's not a limitation on the AI. It's what makes your team comfortable letting it do more.
The AI earns autonomy, task by task.
An assistant starts by drafting. When its drafts stop needing edits, it moves to completing work you approve. When approvals become rubber stamps, it moves to acting with a sample review. Every step up is your call, backed by evidence from your own work.
Your team gets more capable, not smaller.
The goal is never a quieter office. It's your bookkeeper doing the analysis she never had time for, your admin lead running the project she kept putting off, your business absorbing next year's growth without a hiring round.
The honest case against "set and forget".
Fully autonomous AI on business-critical work fails in a specific way: quietly, at scale, and in your name. A wrong invoice reminder sent to one client is a Tuesday. The same reminder sent to two hundred clients overnight is a reputation problem you find out about from your best customer.
The businesses that get burned aren't the ones that moved slowly. They're the ones that skipped the levels. Human-Led AI is how you get the hours back without ever being that story.
Five levels. You choose where each task sits.
| Level | Name | What happens | Where I build |
|---|---|---|---|
| 0 | Manual | A person does the whole task | — |
| 1 | Assist | AI drafts, a person edits and sends | ✔ |
| 2 | Approve | AI completes it, a person approves before it goes out | ✔ |
| 3 | Audit | AI acts, a person reviews a sample afterwards | ✔ |
| 4 | Autonomous | AI acts, nobody checks | Only when earned, and only if you choose |
I build hereLevel 1 — AssistLevel 2 — ApproveLevel 3 — Audit
The foundation is the product. The assistants are what you build on it.
Groundwork is the seven-layer foundation I set up once in your Microsoft 365 tenant: who the AI runs as, what it may see, where it's built, what it may never do, how every action is logged, how new assistants get approved, and the runbook your team uses to build the next one.
It's the reason your first assistant costs what it costs and your tenth costs a fraction. It's also the reason you don't need me afterwards. I publish the whole framework, because the value isn't in knowing what the layers are. It's in having them done properly.
Where the hours go back.
Finance
invoices chased, coding prepared, month-end drafted
Admin & reception
enquiries triaged, bookings confirmed, documents chased
Sales
proposals drafted, CRM kept honest, quiet quotes nudged
HR & people
onboarding packs built, policy questions answered
Operations
reports drafted, clashes flagged, compliance pre-filled
Owners & managers
inbox triaged, board pack drafted, the week summarised
The assistants I build most often.
Invoice and document chasers. Enquiry triage that answers the common questions and hands the unusual ones to a person. Inbox and document summarisers. Proposal and report first-drafters. The copy-paste-between-systems work nobody enjoys.
Each one is described with the spectrum level it usually starts at and the level it usually graduates to.
Five steps from "should we?" to "we do this ourselves now".
1. AI Readiness Review — 1 week.
Where your data lives, who can see it, what your team is already pasting into consumer AI tools, what your subscription already includes. You get a scorecard and a roadmap you keep regardless of what happens next.
2. Groundwork — 3 to 5 weeks.
The seven layers, built in your tenant, with written acceptance criteria for each. Ends with a handover you could show an auditor.
3. First assistant — 2 weeks.
One task, one named owner, one baseline measurement. Built on Groundwork, so it inherits every guardrail without extra work.
4. Team enablement — half a day plus a playbook.
Your nominated person learns to build the next assistant on the same foundation. This is the step that makes you independent of me.
5. Optional: quarterly governance review.
What's been built since, what's drifted, what's ready to graduate a level. Two hours a quarter. Entirely optional, and most clients who take it do so because they've built more than they expected to.
Groundwork
The seven-layer foundation built in your tenant: identity, data boundary, build surface, guardrails, audit trail, change control, runbook.
from $7,500 + GST
3–5 weeks · fixed price
View on pricing →First assistant on Groundwork
One task, one named owner, one baseline measurement.
from $2,200 + GST
2 weeks · fixed price
View on pricing →Team enablement (standalone)
Half-day session plus a written playbook. Included free within Groundwork.
from $1,500 + GST
1 week · fixed price
View on pricing →I don't resell any AI platform or Microsoft licence, and I have no software of my own to sell you. When I say "you don't need AI for this", I mean it. Independence disclosure →
Start wherever you're comfortable.
Low commitment
Fixed price, one week, a scorecard and a roadmap you keep whether or not we go further.
Direct
No pitch. Worst case you leave with two or three things worth fixing.