Three uncomfortable truths about AI in Australian business (Markdown)

## 1,000+ hours of AI research and market analysis, three insights, and what leaders should do with them.

Xperiens Digital launched this month. As part of getting there, I put more than 1,000 hours into AI R&D and market research: talking to leaders, reading the reports, building and breaking things. Three insights kept surfacing. None of them are glamorous. All of them matter to how your business performs over the next few years.

## 1. AI governance is ignored in 90% of businesses, and it's now a key organisational risk

If you've rolled out Copilot, Claude or similar across your organisation, you've made a governance decision whether you meant to or not. Economist Impact research found only 8% of organisations globally maintain a comprehensive AI governance framework, dropping to 2% among small firms. IBM's June 2026 study of 2,000 technology executives backs it up: 77% say AI adoption is already outpacing their governance capability, and two-thirds are accountable for AI systems they don't fully control.

Imagine a world where most businesses had no cybersecurity policy or controls. That was reality not long ago, and we've spent a decade catching up. AI governance is at the same starting line, and the pace of adoption is faster. And the risks are higher as AI-derived "insights" spread into many critical decisions, daily.

The uncomfortable part: this is complex work, and many IT teams and AI service providers don't yet have the capability to do it well. SMEs, enterprises and government departments are all facing the same gap. This isn't a compliance footnote. Ungoverned AI is now a material organisational risk sitting alongside cyber, privacy and financial controls, and it belongs on the same agenda. I've written about the Shadow AI dimension of this here: [Shadow AI: Closing the Gap Between AI Adoption and Governance](/insights/shadow-ai-adoption-governance-gap).

## 2. Tools first, strategy later, value left on the table

A common pattern: the licences were bought, a few pilots ran, some people got clever with prompts, and the organisation is now producing a great deal of AI-generated output that nobody asked for and nobody trusts. Slop, in other words. Please make it stop 😩

I share a fair amount of the concern people have about AI. I also think there has never been a better moment in the history of your business to rethink how it should run. Automation and agentic AI don't just make existing processes faster. They make some processes unnecessary and others possible for the first time. The leaders getting ahead are re-engineering business models and workflows around what intelligent, automated systems can now do, rather than layering tools onto how things have always worked.

But it's still transformation, and transformation succeeds or fails on the quality of the strategy. Relying on non-technical staff to build individual tools and pilots in the hope that value emerges is speculative at best, and it quietly grows the risk footprint across your IT estate. Isolated pilots produce isolated results. Coordinated redesign, with clear ownership and a measurable business case, produces performance. McKinsey's most recent State of AI survey found that AI high performers, the roughly 6% of organisations attributing 5% or more of EBIT to AI, are nearly three times as likely as everyone else to have fundamentally redesigned their workflows, and workflow redesign was one of the strongest predictors of business impact in the whole study.

## 3. "Anyone can build software now" is only half true

AI-assisted development is remarkable. I've built solutions from ideas on a whiteboard to working prototype in days. That part is real.

What's also real is what happens next, and I've seen it play out repeatedly: prototypes that quietly become production systems with no security review; code nobody in the organisation can maintain because nobody in the organisation wrote it; data flowing to third-party models with no record of what went where; integrations that work until the underlying platform changes its pricing or its API; and a growing sprawl of tools with no owner, no support path and no way to switch off.

None of this means leaders should back away. It means they need to understand the new economics of building with AI, including where the costs actually sit and which risks are worth taking. I've covered one piece of that here: [Cutting AI token costs: why open-weight models deserve a place in your stack](/insights/reduce-ai-token-costs-open-weight-models).

## What this adds up to

The organisations that will do well aren't the ones with the most AI tools or the most advanced models or the highest token spend (shakes head). They're the ones that know where AI creates measurable value for them, have redesigned the work around it, and have the governance and capability to scale it without creating unmanaged risk.

That's the work Xperiens Digital was built for. Behind it is 25-plus years of leading businesses through transformation that delivered productivity gains, profit, engaged staff and loyal customers, and a methodology that starts with your business challenges rather than the technology.

**If you're ready to move past experimentation, start by finding out where the value actually is.**

The [AI Value Diagnostic](/services#discover) is a short, high value engagement that gives you a prioritised, costed roadmap, technical brief, and a shared executive view of what to do first. It's primary aim is to show you where the growth opportunities are in your business leveraging frontier technology.

**[Identify your highest-value transformation opportunities →](/contact)**

--- title: "Three uncomfortable truths about AI in Australian business" slug: "three-uncomfortable-truths-about-ai-in-australian-business" date: "2026-09-18" author: "Ben Dexter" topics: ["AI Governance", "Digital Transformation", "Business Strategy"] summary: "1,000+ hours of AI research, three uncomfortable truths, and what Australian business leaders should do about governance, strategy and AI-built software." description: "1,000+ hours of AI research, three uncomfortable truths, and what Australian business leaders should do about governance, strategy and AI-built software." keywords: "AI governance, AI strategy, workflow redesign, AI risk, AI value diagnostic, Australian business" cta: "Talk to Ben about finding where AI creates value in your business" --- ## 1,000+ hours of AI research and market analysis, three insights, and what leaders should do with them. Xperiens Digital launched this month. As part of getting there, I put more than 1,000 hours into AI R&D and market research: talking to leaders, reading the reports, building and breaking things. Three insights kept surfacing. None of them are glamorous. All of them matter to how your business performs over the next few years. ## 1. AI governance is ignored in 90% of businesses, and it's now a key organisational risk If you've rolled out Copilot, Claude or similar across your organisation, you've made a governance decision whether you meant to or not. Economist Impact research found only 8% of organisations globally maintain a comprehensive AI governance framework, dropping to 2% among small firms. IBM's June 2026 study of 2,000 technology executives backs it up: 77% say AI adoption is already outpacing their governance capability, and two-thirds are accountable for AI systems they don't fully control. Imagine a world where most businesses had no cybersecurity policy or controls. That was reality not long ago, and we've spent a decade catching up. AI governance is at the same starting line, and the pace of adoption is faster. And the risks are higher as AI-derived "insights" spread into many critical decisions, daily. The uncomfortable part: this is complex work, and many IT teams and AI service providers don't yet have the capability to do it well. SMEs, enterprises and government departments are all facing the same gap. This isn't a compliance footnote. Ungoverned AI is now a material organisational risk sitting alongside cyber, privacy and financial controls, and it belongs on the same agenda. I've written about the Shadow AI dimension of this here: [Shadow AI: Closing the Gap Between AI Adoption and Governance](/insights/shadow-ai-adoption-governance-gap). ## 2. Tools first, strategy later, value left on the table A common pattern: the licences were bought, a few pilots ran, some people got clever with prompts, and the organisation is now producing a great deal of AI-generated output that nobody asked for and nobody trusts. Slop, in other words. Please make it stop 😩 I share a fair amount of the concern people have about AI. I also think there has never been a better moment in the history of your business to rethink how it should run. Automation and agentic AI don't just make existing processes faster. They make some processes unnecessary and others possible for the first time. The leaders getting ahead are re-engineering business models and workflows around what intelligent, automated systems can now do, rather than layering tools onto how things have always worked. But it's still transformation, and transformation succeeds or fails on the quality of the strategy. Relying on non-technical staff to build individual tools and pilots in the hope that value emerges is speculative at best, and it quietly grows the risk footprint across your IT estate. Isolated pilots produce isolated results. Coordinated redesign, with clear ownership and a measurable business case, produces performance. McKinsey's most recent State of AI survey found that AI high performers, the roughly 6% of organisations attributing 5% or more of EBIT to AI, are nearly three times as likely as everyone else to have fundamentally redesigned their workflows, and workflow redesign was one of the strongest predictors of business impact in the whole study. ## 3. "Anyone can build software now" is only half true AI-assisted development is remarkable. I've built solutions from ideas on a whiteboard to working prototype in days. That part is real. What's also real is what happens next, and I've seen it play out repeatedly: prototypes that quietly become production systems with no security review; code nobody in the organisation can maintain because nobody in the organisation wrote it; data flowing to third-party models with no record of what went where; integrations that work until the underlying platform changes its pricing or its API; and a growing sprawl of tools with no owner, no support path and no way to switch off. None of this means leaders should back away. It means they need to understand the new economics of building with AI, including where the costs actually sit and which risks are worth taking. I've covered one piece of that here: [Cutting AI token costs: why open-weight models deserve a place in your stack](/insights/reduce-ai-token-costs-open-weight-models). ## What this adds up to The organisations that will do well aren't the ones with the most AI tools or the most advanced models or the highest token spend (shakes head). They're the ones that know where AI creates measurable value for them, have redesigned the work around it, and have the governance and capability to scale it without creating unmanaged risk. That's the work Xperiens Digital was built for. Behind it is 25-plus years of leading businesses through transformation that delivered productivity gains, profit, engaged staff and loyal customers, and a methodology that starts with your business challenges rather than the technology. **If you're ready to move past experimentation, start by finding out where the value actually is.** The [AI Value Diagnostic](/services#discover) is a short, high value engagement that gives you a prioritised, costed roadmap, technical brief, and a shared executive view of what to do first. It's primary aim is to show you where the growth opportunities are in your business leveraging frontier technology. **[Identify your highest-value transformation opportunities →](/contact)**