Everyone's using AI. Almost nobody's getting paid for it.

Walk through almost any technology business today and you'll find AI everywhere. Sales is drafting emails with it. Support is summarising tickets. Engineering has a coding assistant, and somebody in finance is quietly using it to tidy up the board pack.

So here's the uncomfortable question. If everyone's using AI, why is almost nobody getting paid for it?

KPMG's October 2026 AI Pulse survey found that 69% of Australian organisations say AI is delivering meaningful business value. Only 2% say adoption has progressed to established ROI.

Sixty-nine per cent think it's working. Two per cent can prove it.

Deloitte's State of AI in the Enterprise research points the same way. Just 28% of Australian respondents have moved at least 40% of their AI pilots into production, and the biggest barrier to getting AI into everyday workflows isn't the technology. It's skills.

The models are remarkably capable. The gap is execution.

Most businesses don't have an AI strategy. They have AI habits.

Look closely at how AI is actually used in most businesses and you'll find a collection of personal habits. One person's clever prompt. Another's workaround for a tedious report. Someone pasting customer data into whichever free tool they found last week.

It looks like progress, and in a narrow sense it is. But none of it is written down, owned or checked, which means it leaves the business the day that person resigns.

AI tends to stall in technology businesses in three predictable ways:

• The hobby problem. Plenty of individual enthusiasm, no shared standard, and no agreed view of what "good" looks like.
• The pilot that never lands. It impresses in the demo, then meets real data, real customers and real accountability. Nobody agreed upfront what result would justify scaling it, so nobody can make the call.
• The "who's on the hook?" problem. If an AI tool put a wrong figure in a proposal tomorrow, or sent a customer something embarrassing, who would find out first, and who would decide what to do? In many businesses the honest answer is "whoever happens to notice".

If any of those sound familiar, you're in good company. You're also leaving money on the table.

Get the order right

Most AI problems are really sequencing problems. That's why the Rev-IQ three-pillar AI strategy framework is deliberately built in order:

1. Educate and empower the workforce.
2. Operationalise AI into how the business actually runs.
3. Productise AI into what you sell.

Think of it like learning to drive. Educate is getting your licence: knowing how the car works, the road rules, and how to drive defensively when other people do unpredictable things. Operationalise is driving every day, safely, in a car that's serviced and insured. Productise is entering the race.

A lot of businesses are lining up for the race before anyone has a licence. It rarely ends well.

Five things the businesses getting results do differently

1. Somebody owns it. Properly.

When AI belongs to everyone, it belongs to no one. The businesses getting results name one senior owner and give them a deliberately short list: three to five use cases at a time, not thirty.

Every candidate is scored on three plain questions:

• Value: if this works, which number moves, and by how much?
• Feasibility: do we have the data, the process and the people to run it next quarter?
• Risk: what's the worst plausible failure, and who does it land on?

They also keep a visible "not now" list. Saying no to a good idea is much easier when you can say "not yet".

2. They train for judgement, not just prompts.

A one-hour all-staff session on "how to use AI" isn't enablement. It's a box being ticked.

What a salesperson needs to know about AI-drafted proposals is very different from what a support agent needs to know about AI-drafted replies, or what an engineer needs to know about AI-generated code. The real skill isn't writing a clever prompt. It's spotting the answer that sounds completely convincing and is completely wrong. AI is very good at confident. It's less reliable at correct.

3. They redesign the work, not just add a tool.

Add AI to a broken process and you get a faster broken process. The businesses getting results pick the workflows that matter, the ones that are repeated, relied on by several people and noticed when quality slips, and they answer four questions in writing:

• Who owns this workflow?
• What does acceptable output look like?
• What's the human review step before anything reaches a customer, a contract or the board?
• What happens when it goes wrong?

In technology businesses, the first candidates are usually support responses, sales outreach and proposals, recurring reporting and code review.

4. They measure what the board actually cares about.

"Hours saved" is a vanity metric until those hours turn up somewhere useful. Before a use case starts, capture the baseline for the number it's meant to move: proposal turnaround, cost to serve, win rate, first-response time, rework.

Then set a decision date. When it arrives, every use case gets one of three outcomes: scale it, fix it, or stop it. That single habit is the most reliable way out of pilot purgatory.

5. They govern before something goes wrong.

Australia's Guidance for AI Adoption, published by the Department of Industry, Science and Resources in October 2025, is a practical place to start. It sets out six essential practices:

• Decide who is accountable
• Understand impacts and plan accordingly
• Measure and manage risks
• Share essential information
• Test and monitor
• Maintain human control

For a small or mid-sized technology business, this doesn't need to be a 40-page policy. A one-page AI policy, an approved tools list, a named owner, a simple register of where AI is used and a clear escalation path will cover most of it. And here's the counterintuitive part: clear rules tend to make people more confident using AI, not less.

Productise last, and test before you charge

Once your own team has real experience running AI day to day, the product questions become much easier to answer honestly. Does this feature solve a customer problem, or are we just following the market? Will customers pay for it, or simply expect it? What does it do to support load, costs and liability?

Internal use cases are cheaper to get wrong and faster to prove. They build the judgement a business needs before it puts AI in front of customers and puts a price on it.

The first 90 days

Days 1 to 30: see it and set the rules. Find out where AI is already being used, by whom and with which tools. Expect a few surprises. Publish a short policy and an approved tools list, name the owner, choose three to five priorities and record their baselines.

Days 31 to 60: build the capability. Run role-specific training for the teams involved. Document the owner, quality standard and review step for each workflow. Launch the pilots with a fixed decision date.

Days 61 to 90: measure and decide. Compare each use case with its baseline. Make the scale, fix or stop call. Report the results to the leadership team or board in business terms, and only then decide which AI product bets deserve real investment.

Five questions for Monday morning

1. Could every member of the leadership team name our top three AI priorities, in order, and give the same answer?
2. Who is accountable for AI across the business, by name?
3. Which AI-assisted work reaches customers today, and what's the review step on each?
4. For each AI initiative, which number is it meant to move, and what was the baseline?
5. If an AI output caused a problem tomorrow, who would find out first and who would act?

If those questions produce long pauses, that's not a failure. It's your starting point.

The question has changed

A few years ago, the interesting question was what AI could do. That question has largely been answered. The question now is what your business does with it.

The businesses that win won't be the ones with the most AI tools. They'll be the ones that treat AI as an operating discipline: owned, measured, governed and built in the right order.

This article is general information only and isn't legal or professional advice.