AI & Automation - AI readiness
Are you actually ready for AI? An honest readiness check.
Before use cases and roadmaps, there is a simpler question: does your business currently have the data, process consistency and ownership that AI needs to work? This page is the readiness assessment for Australian small and medium businesses. Where AI genuinely pays off is covered in our opportunities guide; the strategy engagement is on the AI consulting page.
At a glance
- Who this is for
- Managers and business owners working out whether AI applies to their operation.
- Common finding
- The best first case is usually document handling, not customer-facing chat.
- Our rule
- AI where the input is unstructured; conventional rules everywhere else.
Use cases
Six things AI does well right now.
All six share a trait: unstructured input, high volume, and a cheap way to catch a mistake.
Reading documents so people do not have to
Supplier invoices, purchase orders, delivery dockets, certificates and contracts turned into structured data that flows into your systems for review.
Triaging inbound enquiries
Classifying an email or form by intent, urgency and product, then routing it to the right queue with a draft response attached.
Summarising long threads and calls
Turning a forty-message thread or a site meeting recording into a short summary with actions, so the record survives the conversation.
Drafting repetitive written work
First drafts of quotes, scopes, service responses and reports from your own templates and data, always reviewed before use.
Searching your own knowledge
Answering questions from your procedures, product data and past jobs, with the source document cited so the answer can be checked.
Flagging anomalies for a human
Unusual pricing, duplicate invoices, unexpected stock movement or drifting margin surfaced early - as an alert, not an automatic action.
Getting started
How to trial AI without wasting a quarter.
A narrow, measured trial answers the question far better than a broad pilot that nobody can evaluate.
01
Pick a task with a measurable baseline
Something already counted in hours, volume or turnaround. Without a baseline you cannot tell whether the AI helped.
02
Check the input is available
AI can only read what it can reach. Documents in a shared inbox are workable; knowledge in someone's head is not.
03
Define what 'good' looks like
Agreed accuracy expectations and a sample of correct outputs to test against, drawn from real past work.
04
Put a person in the loop
The first release proposes and a human confirms. Autonomy is only extended to steps that have earned it through measured accuracy.
05
Run it narrow, then measure
One team, one document type, one queue. Compare against the baseline over a real period rather than a demo.
06
Widen or stop
Extend to the next case only if the measurement supports it. Stopping a case that did not work is a successful outcome.
Expectations
Set these before you start.
Most disappointment with AI comes from expectations that were never stated, not from the technology underperforming.
Realistic
What to expect
- Time removed from reading, sorting and retyping
- Faster first response, with quality reviewed by a person
- Better capture of information that used to go unrecorded
- Consistent handling of high-volume, low-judgement work
- Occasional errors, caught by a designed review step
Unrealistic
What to discount
- Guaranteed accuracy without human review
- Replacing a role outright rather than parts of a task
- Useful answers from data the business does not hold
- Value from a tool nobody has been trained to use
- Any specific percentage saving promised before measurement
Responsibility
Three non-negotiables.
01Keep customer and staff data controlled
Where data is processed and retained is a decision made up front, documented, and matched to your privacy obligations.
02Tell people when they are reading AI-assisted output
Internally at minimum. Customer-facing use is disclosed where a reasonable person would expect to be told.
03Never let AI be the only check
Anything with financial, contractual or safety consequence keeps a human approval step regardless of measured accuracy.
Continue exploring
Pick one task and test it properly.
Tell us the task your team finds most tedious and we will tell you honestly whether AI, automation or a process change is the right tool.