Skip to content
Digital Hub

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.

  1. 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.

  2. 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.

  3. 03

    Define what 'good' looks like

    Agreed accuracy expectations and a sample of correct outputs to test against, drawn from real past work.

  4. 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.

  5. 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.

  6. 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.

  1. 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.

  2. 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.

  3. 03Never let AI be the only check

    Anything with financial, contractual or safety consequence keeps a human approval step regardless of measured accuracy.

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.