Things are changing very fast. Some days that feels exciting. Other days it feels like we can't keep up.
We can see possibilities that weren't there before. We can also find ourselves wondering what it all means for our work, our families and the future our children are heading into.
Those feelings can exist together. Being interested in what comes next doesn't mean we're comfortable with everything it might change.
For business leaders, that matters. We're making decisions about new ways of working while the people around us are trying to understand what those decisions mean for them. We may be working through the same questions ourselves.
My view is that we need to give people a clearer way forward. That means being honest about uncertainty, choosing useful problems to solve and making enough room for people to learn.
AI brings these questions into focus. But the starting point is the experience of the person being asked to change.
People can be interested and unsure at the same time
A 2025 University of Melbourne and KPMG global study surveyed more than 48,000 people across 47 countries. It found that 66% intentionally used AI with some regularity, while 46% were willing to trust it.[1]
Use and trust are different things. Someone can try a tool, find it useful and still have reservations about relying on it.
The same study found that only 47% of employees reported receiving AI training. These are global, self-reported findings collected between November 2024 and January 2025; they don't describe every Australian workplace. They do give leaders a reason to ask whether people have the support they need.[1]
I wouldn't assume a quiet team is a comfortable team. A person might be concerned about making a mistake, looking inexperienced or asking a question everyone else seems to understand.
Start by asking what people think the change will mean. What interests them? What worries them? What do they need explained?
Then respond to what they actually say. A general message about opportunity won't resolve a specific concern about someone's role.
Work doesn't sit separately from the rest of life
When we ask someone to change how they work, we're asking a person with responsibilities outside the business as well.
Questions about skills and job security can become questions about household income, children's education and plans for the years ahead. We shouldn't assume everyone is worried about the same things, or ask people to disclose private concerns. We should recognise why a workplace announcement can carry more weight than its wording suggests.
That changes how I think leaders should communicate. Say what you know. Be clear about what you haven't decided. Explain when people will hear more and how they can contribute.
Avoid promising that nothing will change if you can't support that promise. Explain the purpose of the work and the process for making decisions instead.
People deserve a straight answer, including when the answer is that a decision hasn't been made yet.
A convincing demonstration is only the beginning
A demonstration can make a new way of working look simple. The information is ready, the task is clear and the result appears quickly.
A business has more moving parts. Information may be incomplete. Processes may differ between teams. Someone needs to decide whether the output is accurate enough, appropriate for the customer and safe to use.
Research shows why local testing matters. A study published in The Quarterly Journal of Economics in 2025 examined an AI assistant used by 5,172 customer-support agents. It found a 15% average increase in issues resolved per hour. Benefits varied substantially: less experienced workers gained more, while the most experienced workers saw small speed gains and small quality declines.[2]
That is evidence of potential in a specific setting. It is not a forecast for every business or task.
The practical lesson is to test a defined piece of work with the people who do it. Measure the result before deciding how far to extend it.
Start with a problem people recognise
"We need to do more with AI" gives a team very little direction.
"We spend too much time finding the right information before answering a customer" is something people can investigate.
Ask where work becomes frustrating. Where are people copying information, waiting for an answer or correcting the same mistake? Which recurring tasks take time away from customers or more valuable work?
Then understand the cause. The answer might involve AI. It might involve improving a process, organising information or using automation already available in the business.
For an initial trial, choose something useful enough to matter and contained enough to assess. Avoid beginning with decisions that could materially affect someone's employment, health or access to services.
Write the intended outcome in ordinary language. For example: reduce the time spent preparing a weekly internal update while maintaining accuracy and making responsibilities clearer. That gives everyone a shared basis for judging the trial.
Involve the people who know the work
The people doing a job understand details that a process diagram can miss. They know which information is unreliable, which exceptions come up regularly and where a customer needs an experienced person to step in.
Bring that knowledge into the work early. Include someone enthusiastic about the opportunity and someone willing to question it. Both can help produce a better result.
Give people paid working time to learn and test. Adding an experiment to an already full workload makes it harder to understand whether the approach is useful or simply creating more pressure.
Training should use the tasks people actually perform. Show what an acceptable result looks like, how to check it and when to stop and ask for help.
Keep experienced judgement involved. A confident answer can still be wrong, and a faster process can still produce a worse outcome.
Make the boundaries understandable
People need to know what they can use, what information they can share and who is accountable for the result.
Australia's National AI Centre guidance puts accountability, understanding impacts and appropriate oversight into practical adoption guidance. It also recognises that different uses of the same tool can create different risks.[3]
For a first trial, make the working rules easy to find and apply:
- Specify the approved tool, task and information sources.
- Explain which information must stay out of it.
- Name the person responsible for reviewing outputs and deciding whether they can be used.
- Give people a clear way to report mistakes or unexpected behaviour.
- Set conditions for pausing the trial.
These are starting decisions. The controls need to match the work and any relevant obligations. A draft internal summary and a recommendation affecting a person's livelihood need different levels of scrutiny.
Check understanding through examples. A policy people haven't understood won't give them much confidence when they face an unfamiliar situation.
Measure the whole job
An output appearing in seconds tells us very little about the total effort involved.
Include preparing the information, checking the answer, correcting mistakes and dealing with anything that goes wrong. Count the time of everyone involved, including the reviewer.
Use a small set of measures agreed before the trial:
- Total effort: How much working time does the complete task take?
- Quality: How much correction is needed before the output is usable?
- Service: Does the customer or internal recipient get a better result?
- People: Do users feel more capable, and what has happened to workload?
- Cost and risk: Do the benefits justify the running costs, oversight and exposure?
Compare similar tasks and record exceptions. A small trial can tell you whether to keep investigating; it won't establish an annual return with certainty.
Decide what any saved time is for. It might support better customer service, reduce overtime or create space for work that has been neglected. Make that intention visible and check whether it happens.
A practical first 90 days
The following is a suggested starting structure. Adapt the timing to the size of the business, the work and the risks. Some uses will need specialist assessment before a trial can begin.
- Days 1 to 15: Listen to staff, identify current use and choose a business problem. Output: a short problem statement, an accountable leader and a baseline.
- Days 16 to 30: Define the trial, check information and tool suitability, and prepare users. Output: agreed boundaries, training time, measures and pause conditions.
- Days 31 to 60: Run a contained trial with regular feedback. Output: a record of effort, quality, cost, problems and user experience.
- Days 61 to 90: Review the evidence with the team and decide what follows. Output: a decision to stop, improve or extend, with owners and next steps.
Stopping an unhelpful trial is a useful outcome. It gives you evidence before you commit more money or ask more people to change how they work.
If the trial is worthwhile, expand with care. Check whether the next team has the same needs, information and capacity. Carry the learning forward rather than assuming the result will repeat itself.
Leadership means making the next step clearer
We can't remove every uncertainty about the future. We can make better decisions about what happens in our own businesses next.
We can listen before making assumptions. Explain what we're trying to achieve. Give people time to learn. Check whether the promised improvement has reached the people doing the work.
There is plenty to be excited about. People are more likely to find their own reasons for that excitement when they can see a useful purpose and understand their part in it.
You don't need every answer to take a useful first step. You do need to take responsibility for how you take people with you.
Where should your business start?
If this sounds familiar, let's talk about your business.
We're offering a free one-hour consultation with Interfuze to discuss where you are, what's getting in the way and what you'd like to improve.
We're also finishing a standard starting roadmap to help you consider your next steps. Reach out now and you'll be among the first to receive it.
If you want to go deeper, we can scope a separate piece of work to understand your people, processes, information and systems, then build a tailored roadmap for your business with priorities, owners and practical steps.
That gives you a way to move from a broad conversation about change to decisions your business can act on.
Book your free one-hour consultation
Sources and research notes
Research checked 12 November 2026. The recommendations and 90-day structure are the author's practical synthesis, not a framework validated by the studies below.
- University of Melbourne / Melbourne Business School and KPMG, Trust, attitudes and use of Artificial Intelligence: A global study 2025. Survey fieldwork: November 2024 to January 2025, over 48,000 respondents across 47 countries. Figures cited are global self-reported results, not Australian estimates. University research summary
- Brynjolfsson, E., Li, D. and Raymond, L. (2025), Generative AI at Work, The Quarterly Journal of Economics, 140(2), pp. 889 to 942. The final published study reports 5,172 agents and a 15% average productivity increase; earlier working-paper figures differ. The setting was customer support within one company, limiting generalisation. Published paper
- National AI Centre, Guidance for AI adoption: implementation guidance. Practical guidance on accountability, impacts, oversight and other essential practices; it should not be treated as a complete statement of legal obligations. Current guidance
