AI in the Workplace: What Actually Changed

Almost nine in ten organisations now use AI.
Six per cent can find it in the profit line.

‍

Four in five people using AI say it makes them more productive. At organisational level, 37% attribute any positive EBIT contribution to it, and 6% clear McKinsey's high-performer bar of at least 5% of EBIT with an impact they call significant. That 6% has not moved in a year (McKinsey, August 2026).

‍

Adoption finished. The payback hasn't started, and the distance between those two numbers is where this paper sits.

‍

This is a four-page white paper, a 6-minute read, written for senior people leaders deciding what comes after the tooling spend. Written by Graham Dobbin, Director of Change Leadership and Workplace Strategy, Dale Carnegie Australia.

‍

Why isn't our AI investment showing a return?

Because individual gains and enterprise gains have different requirements, and only the first one has been met.

‍

McKinsey's State of AI, published 25 August 2026 from 1,719 respondents across 97 nations, found 80% of people using AI report improved individual productivity, while 37% of organisations attribute any positive EBIT contribution to it and 6% clear the bar of at least 5% of EBIT with a significant impact.

‍

One person changing how they work needs nobody's permission and nothing else has to change. An enterprise return needs a whole workflow to change, which means the people inside it have to understand what was decided, trust it, and feel able to work differently. Those are conditions about information rather than software, which is why the return stalls in organisations that bought well.

Inside the paper

Adoption against payback.

McKinsey's numbers on who uses AI, who gains from it personally, and the small group that can find it in the accounts.

Four measures, one direction.

Confidence, trust, transparency and relevance, each measured separately by level, each falling the same way.

The rework nobody nets off.

Where 40% of the productivity gain goes, and why it never reaches the business case.

Where the next twelve months get decided.

Four levers, with what each one changes and what it costs to run.

Why now

‍

Almost every organisation has bought something. Very few can show what it returned. The shortfall sits in what people were told about the decision and what they were trained to do with the result, neither of which appears on a licence invoice.

‍

Australia is further back than most on the first of those. Fewer than three in ten workers say their organisation is clearly aligned on AI strategy, and only a third have had any formal training. That combination produces exactly the pattern the global numbers describe: real individual gains, plenty of rework, and nothing visible at enterprise level.

‍

Key numbers and ideas

of organisations attribute at least 5% of EBIT to AI with an impact they call significant, flat on the previous year

34% v 9%

leaders who feel confident adapting to AI-driven change, against individual contributors

Every gap in this data widens in the same direction. Down.

‍

The technology arrived. The explaining did not.

‍

About the research

The paper draws on McKinsey's The State of AI, published 25 August 2026 (1,719 respondents across 97 nations); Dale Carnegie's AI in the Workplace: Navigating Generational and Role-Based Perceptions (3,375 respondents across 18 countries, research conducted 2024, white paper published January 2025 and cited here by publication year); Workday's April 2026 research; Microsoft's 2026 Work Trend Index Australian findings (2,000 Australian workers within a global sample of 20,000 across 10 countries); EY's Australian AI Workforce Blueprint, 2025 (1,003 Australian workers who use computers daily); and the Hays Salary Guide FY26/27 (more than 7,000 professionals across Australia and New Zealand combined). Written by Graham Dobbin, Director of Change Leadership and Workplace Strategy, Dale Carnegie Australia. Dale Carnegie global research is led by Robert Coleman, PhD, Director of Research and Thought Leadership.

About Dale Carnegie

Dale Carnegie has been studying what changes how people communicate, lead, and influence at work since 1912. Over 9 million graduates across 90+ countries. In Australia, we work with HR and people leaders at organisations including BHP, Macquarie, GE Healthcare, Schneider Electric, and Veolia to turn capability frameworks into behavioural change you can see in the room.

Download the
Insight Report

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

How much of the AI productivity gain is lost to rework?

About 40%, according to Workday research published in April 2026. Roughly four hours come back for every ten saved.

‍

The saving is usually measured at the point of generation and never netted off against the correction, so it never reaches a set of accounts however real it felt to the person who made it. The rework lands on a different desk, in a different week, under a different heading.

‍

Any AI business case that counts hours saved without counting hours returned is measuring half the transaction.

‍

Why won't employees use the AI tools we've bought?

Because confidence, trust, transparency and relevance all fall sharply with seniority, and each shortfall creates the next one down.

‍

Dale Carnegie's 2025 study of 3,375 employees across 18 countries measured four things separately and split every answer by role. Confidence in adapting to AI-driven change: 34% of leaders, 20% of managers, 9% of individual contributors. Whether AI has directly affected their own role: 24%, 12% and 4%. Trust in leadership's AI decisions: 37% against 9%. Perceived transparency: 31%, 18% and 6%.

‍

Transparency is what people can see about how a decision was made. Trust is whether they believe it was sound. Confidence is whether they feel able to work differently. Relevance is whether any of it reached the actual job, and barely one contributor in twenty says it has.

‍

The sharpest figure in the set is not about the workforce. 63% of leaders do not have a high level of trust in the leaders above them to get AI decisions right.

‍

Is resistance to AI a generational problem?

No. What falls across the generations is exposure rather than aptitude.

‍

Confidence in adapting to AI-driven change runs at 25% among Gen Z, 21% of Millennials, 13% of Gen X and 8% of Boomers (Dale Carnegie, 2025, 3,375 employees across 18 countries). Australian self-rated AI proficiency follows the same descent, scoring 46, 37, 25 and 18 out of 100 across those four groups (EY, 2025, 1,003 Australian workers).

‍

What descends across those groups is contact hours. The people at the top of the list have used these tools in study, in side projects, and in jobs where nobody was watching closely enough to make a mistake expensive.

‍

That distinction matters commercially. An organisation cannot change the age profile of its workforce. It can change how much practice people get.

‍

What percentage of Australian workers have had AI training?

35%, according to EY's 2025 Australian AI Workforce Blueprint, a survey of 1,003 Australian workers who use computers daily.

‍

The same research found 54% are not confident using AI in their work, 42% have not been given a clear reason to use it in their own role, and two thirds want their employer to provide more.

‍

Hays measured it from the other end in its FY26/27 Salary Guide, across more than 7,000 hiring managers and professionals in Australia and New Zealand: 60% use AI regularly at work, and 78% of those said their employer had provided no formal training at all. The two are not directly comparable, because Hays surveyed people already using AI and combined the Australian and New Zealand samples.

‍

Microsoft's 2026 Work Trend Index adds the strategic half, from 2,000 Australian workers: 28% say their organisation is clearly aligned on AI strategy, 51% say it feels safer to focus on current goals than to rethink how they work, and 13% say reinvention gets recognised when results take time.

‍

How do you get a return on AI already bought?

Four moves, none of which requires a new platform or a bigger licence.

Measure the gap by level before buying anything else. An organisational average hides a spread running from 34% to 9%. Ask leaders, managers and the people doing the work the same four questions, then report the three answers separately rather than blending them into one number that describes nobody.
Explain the decision to the people who have to live with it. Say what was decided, why, and what changes for each team. Transparency tracks trust throughout the data and it is the one measure a leader can move this quarter.
Give people a reason for their own job. 42% of Australian workers say they have not been given a clear purpose for using AI in their role (EY, 2025). A company-level strategy statement does not answer that question for the person doing the work.
Count the rework before counting the saving. If 40% of the gain goes back into correction (Workday, 2026), a saving nobody has netted off is not yet a saving.

More Resources

View all

Insight Report

Rapport Gets the Meeting: It Doesn't Keep the Room

When answers are cheap and everyone can be creative, the advantage sits with the people who can read a client's business, ask the question nobody has asked, and make a case worth defending.

Insight Report

Critical Thinking: The Capability We Skipped

Employers have named analytical thinking their number one skill for years. One employee in five has had any development in it.

Your Cart

AI in the Workplace: What Actually Changed