Why technology adoption isn’t translating into business value…and what leaders need to redesign 

It’s hard to avoid AI at the moment. 

Most organisations have moved beyond the initial debate about whether AI matters. Licences are being bought. Pilots are underway. People are experimenting. Adoption is increasing. 

You can sense the potential for the technology to change how organisations work, hopefully for the better. 

The problem is that the measurable business impact is proving harder to find. 

PwC’s 29th Global CEO Survey reports that only 12% of CEOs are achieving both cost and revenue benefits from AI, while 56% report neither. 

Deloitte identifies a similar disconnect. Nearly 60% of workers intentionally use AI at work, yet relatively few organisations believe they are genuinely good at designing how humans and AI work together. 

We would argue that this gap isn’t primarily an AI capability problem. 

It’s a leadership-design problem. 

Many organisations are still managing AI much as they would manage a conventional technology implementation: 

Buy the licences. Train people. Encourage adoption. Measure usage. Wait for value to appear. 

(To be fair, our historical record of extracting value from major technology investments has always been somewhat mixed.) 

But AI presents a different challenge. 

It doesn’t simply change the tools people use. It changes how work gets done, who makes decisions, where accountability sits, how people develop expertise and, increasingly, what we expect managers and leaders to do. 

AI adoption might reduce costs initially however it doesn’t create value because people use it. It creates value when leaders redesign work, judgment and accountability around it. 

We may be making a category error 

Deloitte’s 2026 Global Human Capital Trends research found that 59% of organisations are taking a predominantly technology-focused approach to AI. 

Those organisations are 1.6 times more likely to fall short of expected returns than organisations taking a more human-centred approach. 

Yet the instinctive response of many leadership teams is still technological. 

Do we need better models? 

Better data? 

Better integration? 

More training? 

A different platform? 

Sometimes the answer will be yes. But increasingly we may be looking in the wrong place. The first question at the executive table shouldn’t be: 

Where can we deploy AI? 

It should be: 

How should this work now be designed? 

That is a much more demanding leadership question. 

AI changes work, not just tasks 

Much of the conversation about AI has focused on tasks. Which tasks can AI automate? Which can it accelerate? Which jobs contain the most automatable work? 

Those are useful questions, but incomplete ones. 

Because when AI absorbs part of a role, the remaining human work changes as well. 

Imagine a manager who previously spent several hours analysing performance information, preparing reports and developing recommendations. 

AI can increasingly produce much of that analysis in minutes. That sounds like a significant productivity gain.

•  But what happens next?
•  What does the manager do with the time released?
•  And perhaps more importantly, what capabilities does that manager now need?

The value of the role may increasingly lie in interrogating the analysis, spotting what AI has missed, understanding context, balancing competing priorities, having difficult conversations and exercising judgment when the evidence is incomplete. 

Removing lower-value cognitive work doesn’t necessarily make human capability less important. 

It can make the remaining human capability more important. 

This is why leaders and managers matter so much. The executive team may approve the AI strategy, but managers will determine whether it works. They will translate broad policies into hundreds of everyday judgments:

•  Can I trust this analysis?
•  Should I check it?
•  Can this go directly to a client?
•  When does a human need to intervene?
•  What does good performance look like now?

If managers aren’t equipped to answer those questions, the organisation doesn’t really have an AI operating model. 

It has a collection of individual experiments. 

The dangerous productivity illusion 

AI also creates the possibility of a new kind of productivity illusion. 

It doesn’t necessarily level performance across a workforce. It can amplify the capability already there. 

Experienced people who understand their domain, recognise weak assumptions and know which questions to ask can become considerably more effective with AI. 

But less-experienced people can also produce something remarkably convincing. 

AI changes that relationship. 

Weak reasoning can now arrive beautifully structured, professionally written and apparently authoritative. The slides look good. The grammar is excellent. There are headings, tables and recommendations. 

But the reasoning may still be wrong. 

That shifts the leadership challenge from reviewing output to evaluating judgment. 

And there is another productivity illusion that may be harder to see. 

Call it the AI verification tax. 

Suppose AI saves someone 30 minutes producing an analysis. They then spend 20 minutes checking it because they aren’t sure how much they’re supposed to trust. 

They rerun the numbers. Ask a colleague to verify the answer. Or quietly do the work the old way as well, just to be safe. 

On the adoption dashboard, AI is working. In reality, much of the productivity gain has disappeared. That isn’t necessarily resistance. It is rational behaviour inside an ambiguous system. 

When people don’t know where the boundary sits between AI judgment and human judgment, they often respond by checking everything. Or, potentially more dangerously, checking nothing. 

Clear decision rights aren’t just about governance. They help release the productivity gain AI was supposed to create. 

There is a longer-term issue too. 

The expertise paradox 

Junior people have traditionally developed expertise by doing many of the tasks AI is beginning to absorb: researching, drafting, analysing, summarising and solving relatively straightforward problems. 

Those tasks weren’t simply production. They were practice. 

The junior analyst who spends hours constructing an analysis learns how the numbers fit together.  The new consultant drafting the first version of a report learns how to structure an argument. The emerging manager working through a difficult problem develops judgment by getting things wrong, receiving feedback and trying again. 

If AI increasingly performs that work, what replaces the learning? 

Organisations risk creating an expertise paradox: today’s productivity gain becomes tomorrow’s capability shortage. Which leads to an important question: 

If AI does the work through which your experts learned to become experts, how will your next generation become expert? 

The answer is not to preserve low-value work simply because that is how people used to learn. It is to redesign the apprenticeship pathway deliberately. 

Leaders need to think not only about what AI can remove from a role, but what experiences people still need in order to develop judgment, expertise and professional intuition. 

That isn’t an IT question. It’s a leadership development question. 

So, what is the human advantage? 

The human advantage isn’t competing with AI at things AI increasingly does well. Humans don’t need to beat AI at producing a first draft in 20 seconds. 

The advantage lies elsewhere.

•  Understanding context
•  Asking better questions
•  Navingating competing values
•  Building relationships and trust
•  Recognising when something doesn’t feel right despite a convincing answer
•  Taking responsibility for a decision
•  Exercising judgement where there is no obviously correct answer

McKinsey describes an important shift in leadership as AI takes on more analytical and execution work: from leaders primarily providing commands to leaders providing context; guardrails, priorities and the judgment needed for humans and AI to work effectively together. 

That is a significant change. 

It asks leaders to move away from the apparent certainty of control towards something more difficult: judgment under ambiguity. 

And as organisations increasingly gain access to similar AI technologies, this becomes strategically important. 

The differentiator won’t simply be who has AI. 

It will be who has redesigned the organisation to use it well. 

Three things leaders need to redesign 

The leadership challenge can be simplified into three areas. 

1. Redesign the work

Don’t simply bolt AI onto an existing role. 

Ask: 

•  What work should disappear?
•  What should be automated?
•  What should be augmented?
•  Where should human time and attention now be concentrated? 

Then redefine what good performance looks like. 

And don’t forget development. If AI removes the work through which people previously learned, deliberately create new ways for them to build expertise.

2. Redesign the decisions

Every significant AI implementation should define its decision boundaries. 

A useful starting point is: 

AI decides. 

AI recommends; human decides. 

Human only. 

Not every decision requires a governance committee. 

But consequential decisions need clear ownership. 

Otherwise organisations can outsource judgment almost by accident, and discover during the first serious problem that nobody knows who was accountable.

3. Redesign the leadership

Managers need more than AI literacy. 

They need to be able to provide context, challenge AI-generated thinking, coach people whose work is changing, manage the boundary between trust and verification and exercise judgment when AI provides an answer but not necessarily the answer. 

They also need to create sufficient psychological safety for people to challenge AI-generated conclusions, question assumptions and raise concerns rather than simply accepting apparently authoritative output. 

This may be the biggest capability challenge of all. 

Because the future of AI inside organisations won’t be determined solely by what the technology can do. It will be determined by thousands of decisions managers and leaders make about when, where and how it should be used. 

Technology plan or leadership plan? 

Here is a simple test. 

Take one of your major AI initiatives and ask: 

If we switched the AI off tomorrow, apart from doing things more slowly, what would actually be different about how this role operates? 

If the answer is not much, you probably haven’t redesigned the work. 

Then ask five questions. 

If most of your answers sit in the middle column, you probably don’t yet have an AI strategy. 

You have a technology programme wearing strategy’s clothing. 

A particular opportunity for New Zealand 

The New Zealand numbers make the issue more pressing. 

Deloitte’s 2026 research found that only 2% of New Zealand organisations describe themselves as leading in intentionally designing how humans and AI work together, compared with 7% globally. 

More tellingly, 55% of New Zealand respondents identify insufficient understanding as their biggest challenge in addressing AI’s implications for decision-making and leadership. 

Being behind clearly has risks. 

But it also creates an opportunity. 

A particular opportunity for Australia 

Australia faces a similar challenge. Deloitte’s 2026 State of AI in the Enterprise report found that while organisations are investing heavily in AI, many are still struggling to translate adoption into genuine transformation. Only 12% of Australian leaders reported that generative AI is already transforming their business or industry, compared with 25% globally. While 61% reported productivity improvements, only 30% said they were using AI to deeply transform how work is done. Deloitte argues that many organisations remain focused on automating existing processes rather than reimagining work itself. 

The message is clear: the opportunity for Australian leaders is not simply to use more AI, but to redesign work, decision-making and organisational capability so that AI delivers lasting business value.

Where to start 

Pick one live AI implementation where adoption is increasing but the business value remains unclear. 

Put the technology aside for an hour and ask: 

•  What work has actually changed?
•  What should people stop doing?
•  Where should AI recommend rather than decide?
•  Where is human judgment non-negotiable?
•  Who owns the final decision?
•  How will people develop expertise?
•  What business outcome should improve if this is genuinely working? 

Then change one thing. 

Redesign one role. 

Clarify one decision right. 

Name one accountable owner. 

Remove one unnecessary verification step. 

Replace one adoption metric with an outcome metric. 

Then come back thirty days later and ask whether anything meaningful has changed. 

Access to powerful AI will increasingly become commonplace. 

The divide will be between organisations that simply use AI and those that have learned how to lead differently because AI exists.

Like to know more about how to use practical neuroscience to drive more impact?

Register for one of our public leadership programmes

Connect with us on LinkedIn, or sign up to our e-newsletter