orangeAI
Discover · phase 1
AI Readiness Scorecard FreeA three-minute self-check.AI Opportunity Call FreeA straight steer on your best next step.AI Activation $2,499Switched on as a client, map in hand.AI Discovery & Roadmap AnchorKnow exactly where AI earns its place.Process CaptureHow the work happens, documented as an asset.Reasoning CaptureKey-person judgement, made an asset.
Design & Ignite · phases 2–3
Design & BuildThe engine that constructs it all, including skill design.
Amplify · phase 4
Chief AI OfficerAn accountable AI executive on retainer.Managed AgentsHigh-stakes work, run as a governed service.Business IntelligenceLive cockpits, assembled and analysed overnight.Governed AI RolloutYour whole workforce, switched on safely.
One methodology: Discover → Design → Ignite → Amplify. Every product sits on it.How I work →The Client Portal →Full map →
MethodologyCase studiesInsightsAboutClient Portal ↗
Free AI Scorecard
Home
Discover
AI Readiness Scorecard · FreeAI Opportunity Call · FreeAI Activation · $2,499AI Discovery & RoadmapProcess CaptureReasoning Capture
Design & Ignite
Design & Build
Amplify
Chief AI OfficerManaged AgentsBusiness IntelligenceGoverned AI Rollout
Company
MethodologyHow I workCase studiesThe Client PortalInsightsAboutClient Portal loginContact us · hello@orangeai.com.auStart here: free AI Scorecard or a call
orangeAI

We map how your business actually runs, then put AI where it earns its place. Brisbane, Australia.

What we do
What we doMethodologyCase studiesInsights
Company
AboutContactPortal login
Follow

Content lands on LinkedIn first.

The Lens, weekly
PartnersMicrosoft PartnerAnthropic · Claude Partner Network
Orange AI Pty Ltd (ABN 55 697 856 447) · Privacy · Terms · Contact© 2026 Orange AI Pty Ltd
BlogAI AdoptionAI ImplementationOperating Model

Your Staff Adopted AI. Your Business Didn't.

By Brad Ferris · 26 September 2026

5 min read

On 24 September the Australian Financial Review published an opinion piece by Microsoft ANZ president Jane Livesey, I run Microsoft Australia. Here's the AI number keeping me up at night. It puts two numbers side by side. Microsoft's own data ranks Australia 11th in the world, with 40.7 per cent of working-age Australians using generative AI in June against a global average of 18.8 per cent. Treasury, meanwhile, reports that fewer than one in ten Australian businesses would call their own adoption of AI significant.

Read those together and a common diagnosis falls over. The usual story in mid-market businesses is that Australians are cautious, that staff need convincing, that adoption is a training problem. The staff have moved. Somewhere between the individual and the organisation, the momentum stops.

Where the momentum stops

Livesey describes a first wave of adoption, counted in licences, pilots and hours saved. Most operators will recognise it. A handful of people use a chat tool every day. Someone in finance has a prompt that reconciles supplier statements in half the time. Someone in sales drafts every proposal with it. The leadership team knows this is going on in a general way and could name perhaps a third of it.

Very little of that shows up in the numbers the business runs on. The hours saved are real, and they get absorbed into each person's week without anyone changing a process, a role or a target. The capability belongs to the individual. When that person leaves, it leaves with them. When they get something wrong, nobody else is positioned to notice.

That is the shape of the gap. Individual use grows on its own momentum. Organisational use needs somebody to decide which work changes, write the new method down and answer for the result. In a business of 50 to 300 people, that job usually has not been given to anyone.

The tools will not close it for you

Ethan Mollick made a related argument in The Overhang on 18 September. Current models can already do weeks of human work when properly guided, he writes, yet their capabilities are barely being used, because human systems and processes move slowly. On that reading, the scarce input in most businesses is direction: somebody deciding what the tools should be pointed at.

The largest workplace software vendor appears to have reached a similar view about where value sits. Bloomberg, reported via Business Standard, says Microsoft is merging the consumer and workplace versions of Copilot into one product aimed at corporate customers, ceding the personal chatbot market to OpenAI, Google and Meta, and letting users edit Word and Excel documents from inside the assistant. Whatever you make of the product, the direction of travel is clear: the value is in work the business already does, inside software it already runs.

That should be reassuring for an owner who feels behind. The discovery work has largely been done by your own people, for free, on their own initiative. What is missing is the step that turns their habits into the company's method.

Four moves that turn personal use into business capability

  1. Count what is already happening. Before buying anything, ask. A short anonymous survey, or ten conversations with team leads, will surface which tasks staff already hand to AI and which tools they use to do it. Treat the answers as a map of demand rather than a list of breaches. Punishing the early adopters guarantees the next discovery happens out of sight.
  1. Pick one workflow and make it the company's. Choose a task where several people have independently landed on AI: quoting, supplier reconciliation, first-draft customer replies. Write down the steps, the prompt, the checks and who reviews the output, and put it somewhere shared. Once it is documented, it survives staff turnover and more than one person can improve it.
  1. Use the simplest tool that does the job. The ATO is a useful counterweight to the rush toward agents. Assistant commissioner Jimmy Tzimopoulos told [iTnews](https://www.itnews.com.au/news/ato-culture-keeps-a-cap-on-agentic-ai-ambitions-629118) that when the ATO assessed its use cases, it might find 95 per cent could be achieved through process re-engineering, business automation and some machine learning, with agentic AI as a component at the end. The ratio in your business will differ, but the order of operations holds: fix the process, then automate it, then decide where judgement from a model earns its place.
  1. Give it an owner and a number. Each sanctioned workflow needs a senior person accountable for it and a measure the business already tracks, such as turnaround time, error rate or cost per transaction. Without the number, the hours saved stay invisible to the P&L. Without the owner, the workflow drifts back to whoever happened to care most.

The trust figure cuts both ways

The headline of Livesey's piece is a different statistic again. Only 1 per cent of Australians fully trust AI, and 44 per cent have little or no trust in it, which leaves the country using the technology more than most while trusting it less than almost anyone.

High use combined with low trust is the profile of quiet, unsanctioned adoption. People reach for a tool that saves them time, keep it to themselves, and check the output only as carefully as their own judgement demands. That is the risk most owners worry about: something wrong goes out under the company's name, and nobody knew AI was involved.

Bringing that use into the open, with a documented method and a named reviewer, deals with the risk and the trust problem in the same motion. Staff trust a process they can see others checking. Customers trust a business that can explain how its work was produced.

The part of Livesey's argument worth taking to your next leadership meeting is the simplest. Australia is not short of access to AI, and neither is your business. Your people have already found where it helps them. The job left for the leadership team is deciding which of those discoveries the business will own.


Where does your business stand? The free AI Scorecard takes three minutes and shows you. If you want a straight steer from a person, book an AI Opportunity Call.

Sources
  • I run Microsoft Australia. Here's the AI number keeping me up at night · Australian Financial Review
  • The Overhang · One Useful Thing (Ethan Mollick)
  • Microsoft abandons personal AI chatbot race with Copilot reboot · Business Standard (Bloomberg)
  • ATO culture keeps a cap on agentic AI ambitions · iTnews
What's your next move?

Find out where your business sits on the AI maturity curve.

Take the free AI ScorecardBack to insights