Least-privilege access
Each AI Employee gets only the permissions its job needs, through your own admin console, revocable at any time.
Onboarding in 4 weeks or less
From “we are drowning in admin” to a working AI Employee
No black box. Here is exactly what happens, what we need from you, and how we keep an AI teammate safe to run inside a real business.
The process
Six stages, one painful process at a time
We deliberately refuse to automate ten things at once. One role, proven, beats a platform nobody trusts.
Week 0 — Free AI audit
A 30-minute conversation about where the hours go. We map your repetitive tasks, score them on volume, risk and effort, and send you a one-page plan naming the first role to hire and the outcome we will measure. No obligation, and yours to keep.
Week 1 — Configure the job scope
We write your AI Employee's job description with you: what it does, what it must never do, approval thresholds, tone of voice, escalation rules. This document is the contract between your business and the AI — and you own it.
Week 2 — Connect and train on your data
We connect your systems with least-privilege access (Xero, Google Workspace, WhatsApp Business, your CRM), then run the AI Employee on historical work — last quarter's bank lines, last month's applicants — and compare its output against what actually happened.
Week 3 — Go live, draft-first
It starts working on live volume, but nothing external is sent without approval. You review a daily queue; our Singapore reviewer checks accuracy and tunes rules. Most clients are approving in batches within days.
Week 4 onwards — Widen autonomy
As accuracy proves out, you release the handbrake category by category: standard replies auto-send, routine postings auto-post, reminders go out unattended. You keep the log, the limits and the off switch.
Ongoing — Review and expand
A monthly review covers what it handled, what it escalated, hours saved and errors caught. When the first role is boring in a good way, we scope the next one.
Your side of the deal
What we need from you
Less than you think, but it cannot be zero. The businesses that get the most out of an AI Employee treat onboarding like they would a new hire's first fortnight.
We handle the technical work, the prompt engineering, the integrations and the monitoring. You supply context and decisions.
Access to the systems involved
Read-only first, write access only where needed. We use your admin console so you can revoke us in one click.
Two hours of your time in week one
One scoping session and one rules review. After that, 15 minutes a day reviewing a queue is typical.
Examples of good work
Past emails, a good month of books, your handbook, three posts you liked. Examples teach faster than instructions.
A named decision maker
Someone who can say “yes, that is how we do it”. Usually the owner, finance manager or ops lead.
Safety & governance
How we make an AI teammate safe to run
The technology is the easy part. Controls are what let a business owner sleep.
Everything is logged
Every read, draft, send and posting is recorded with a timestamp and the rule that triggered it. Exportable for audit.
Hard guardrails
Payment and value thresholds, blocked recipients, forbidden claims, topics that must always escalate to a human.
Named human reviewer
A Singapore-based specialist owns quality for your account and is reachable on WhatsApp during business hours.
Reversible by design
Drafts before sends, staged postings before finalisation, and a documented rollback for anything that reaches a customer.
No training on your data
Your documents and correspondence are never used to train public models. Your context stays in your workspace.
What we will not do
- Submit statutory filings without a qualified human reviewing them
- Give legal, tax or employment-law advice — we prepare, professionals approve
- Message contacts who have no lawful basis or have opted out
- Make binding pricing or contractual commitments on your behalf without written rules
- Handle grievance, disciplinary or termination conversations without a human leading
FAQ
Onboarding questions
Practical answers about time, effort and control.
Most clients see net time savings in week 3 — earlier for inbox and enquiry work, a full cycle (one month-end) for accounting. Weeks 1 and 2 cost you a couple of hours of setup.
Tell your reviewer, or update the rules document. Changes take hours, not a change-request process. This is why we keep the job scope as a living document you own.
For most SMEs our managed Singapore-region environment is the pragmatic choice. If you have specific data-residency or client-contract requirements, we can discuss a dedicated deployment.
Thirty days' written notice, no exit fee. You keep your data, your rules document and your logs. We help hand the process back to your team or your next provider.
Get started
Book your free AI audit
Book a free 30-minute AI audit. We map the repetitive work in your business, show you what an AI Employee would take over, and give you a written plan — whether or not you hire us.