Only 2 Per Cent Can Show AI ROI. The Gap Is a People Problem.

Seventy-five per cent of Australian organisations say AI will remain a top investment priority. Only 2 per cent say their adoption has progressed to established ROI.

Both figures come from KPMG's latest AI Pulse survey, whose Australian insights were released in October. They match a pattern that's common across the market. The budget is there. The licences are there. What's missing is someone whose job it is to turn all of that into a result.

What the gap actually tells you

KPMG's own read is blunt: adoption is moving faster than value. KPMG Australia's Anthony Coops puts it well. Scaling AI "requires redesigning work, not simply rolling out new tools."

I agree, and I'd go one step further. Redesigning work is a people job, which makes it a hiring and role design job.

The same survey shows where the pressure sits. Eighty-three per cent expect rising AI fluency to reshape roles. Seventy-nine per cent say employees who use AI effectively outperform others. And 52 per cent say value generating use is their biggest workforce challenge.

That's not a technology problem. It's an ownership problem.

The roles that close the gap

When AI stalls, it's rarely because nobody bought the right platform. It's usually because nobody owns the space between the tool and the work. Three roles keep coming up.

Someone who owns adoption. Not just running training sessions, but changing how a team actually works week to week. I've set out what that looks like in what an AI enablement manager actually does and in my AI Adoption Lead job description.

Someone who can build and run agents properly. KPMG found 42 per cent cite agentic architecture skills as a barrier. SEEK's September Employment Report shows the same squeeze from the hiring side: mentions of agentic AI skills in job ads have nearly tripled in a year, up 199.2 per cent. That's a small pool with a lot of employers fishing in it. If you're budgeting for one, check my generative AI engineer salary guide before you go to market.

Someone who measures value. Only 58 per cent of Australian organisations report visibility of their total AI cost. If nobody can see the cost, nobody can show the return.

My guide to every role in an AI team shows how these fit together and the order most businesses hire them in.

If you're hiring into this

Pick fewer use cases. KPMG's own priority is to convert AI adoption into measurable value. Two or three done properly will teach you more than twenty pilots.

Name the owner before you name the tool. Every use case needs one person accountable for whether it delivers. If that person doesn't exist yet, that's your first hire, not your next licence.

Decide what to build and what to buy. In the Bot, Build, Borrow, Buy framework, adoption roles are often a build moment, because the people who already know your workflows can grow into them. Agent architecture is usually a borrow or a buy, because the skills are scarce and the timeline matters.

Where this is going

The businesses that show a return on AI next year won't be the ones that spent the most on it. They'll be the ones that decided early who was accountable for making it work.

Buying AI is the easy part. Getting a return on it is a people decision.