Key points

• An AI agent doesn’t just answer. It goes off and does the steps on your behalf.
• The same capability that books your trip can copy a voice, a face or a bank notice well enough to fool you, and the scam scripts are written to reach your heart before your head.
• Australians reported $2.18 billion in scam losses in 2025. The defence that works costs nothing: a pause.
• Rules written out of fear tend to push behaviour out of sight rather than stop it.
• On 13 October I launched Choose Forward, an agent built to ask what you’re really asking before it answers.

Phil Whelan opened Tuesday’s segment by asking whether he was talking to me or to an AI agent. We both laughed.

It’s a joke with a shorter shelf life than either of us would like. Three weeks ago Netflix premiered a reality show narrated by Gene Wilder, who died in 2016. His voice was rebuilt with AI, with his family’s blessing, and it sounds like him.

The question Phil asked for a laugh is one plenty of people will soon be asking for real, on the phone, about someone they love.

This week on RTHK Radio 3 Morning Brew, Phil Whelan and I talked about AI agents. I’d come in with a list of where agents are already at work, and we got through about a minute of it before the chat went somewhere more useful.

It went to what agents do, what the same skill does in the wrong hands, why fear makes bad rules, and the thing I’d switched on a few hours earlier.

What an agent actually does

Most people still meet AI as a chatbot. You type a question, it answers, and it seems clever, because it is.

An agent is the next step up. Think about a travel agent. You don’t just ask where to go.

You tell them you want somewhere warm, a beach, a couple of good dinners, and they come back with flights, hotels and a day-by-day plan. If you’re comfortable with it, they book the lot.

An AI agent works the same way. You hand it a job and it does the steps: the searching, the comparing, the forms, the follow-ups. Nothing gets booked until you say so, unless you’ve told it otherwise.

I wrote last week about where these agents have actually turned up, from funeral homes answering the phone at three in the morning to small businesses chasing unpaid invoices. That piece is about the work. This one is about the trust that work depends on.

The same skill, pointed the other way

An agent that can act on your behalf can also be made to sound like someone acting on theirs.

The Gene Wilder show is Wonka’s The Golden Ticket, a competition series that premiered on Netflix on 23 September. His voice was recreated with ElevenLabs, with permission from his widow, Karen, and the estate. As I said to Phil, when you stop and think that he’s gone and he’s still voicing the show, it’s spooky.

It’s also the respectable end of the technology. Somebody asked. Somebody said yes.

The other end doesn’t ask anybody. Voices can be cloned from short samples. Video can be made to look like you or me.

Bank notices, the little ones we’ve learned to recognise because we know where they come from, can be copied at speed and with a confidence that makes the copy harder to tell from the original than the original ever was to read.

I wrote last month about how reputation has become something that can be forged, and the well-known Australians whose faces and voices were spent by other people. What came up on air this week was the private version of that story, the one that lands on an ordinary phone at an ordinary hour.

You know the script. Mum, I’ve lost my phone. I’m stuck. I’ve borrowed someone’s to message you. Can you send some money? It’s written psychologically.

It wants an answer now. And your heart goes out a little before your brain does, which is exactly what it’s designed for.

The Australian numbers say this is not a niche problem. The National Anti-Scam Centre’s Targeting Scams Report for 2025 puts combined reported losses at $2.18 billion, up 7.8 per cent on 2024. Phishing was the most reported scam type to Scamwatch, with 65,361 reports.

Australians aged 65 and over make up about 17.1 per cent of the population and accounted for 26.5 per cent of the losses reported to Scamwatch.

Those are the losses people reported. Plenty don’t, because being fooled by your own child’s voice is not something people rush to tell anyone.

The pause is the defence

What I said on air is the same thing I said about social media years ago. If you wouldn’t stand in the local market and yell it, don’t post it. The equivalent for this moment is simpler still.

If a message asks for money or details in a hurry, the hurry is the warning.

Take a step back. Ring the person on the number you already have. Ring the bank on the number you looked up yourself, never the one in the message.

The National Anti-Scam Centre’s own advice runs to three words, Stop. Check. Protect., and the first of them is the one that does most of the work.

There’s one more check I suggested to Phil’s listeners, and it surprises people. Paste the message into your own AI and ask whether it’s real. It has no skin in the game.

More often than not it’ll tell you this is a known trap, and even when it isn’t sure, it’ll tell you to ring the bank directly. Treat it as a second opinion, never the final word. The decision stays with you.

For three years the pitch for AI has been speed. Faster answers, faster drafts, faster everything. In this corner of life, the most protective thing the technology can do for you is slow you down.

Why fear writes bad rules

Phil asked why I’d rather keep governments at arm’s length from AI. My answer was about people more than governments.

When something frightens us, we overcompensate. The rule arrives as a knee-jerk: this will solve it, this is the only way, and it has to happen now. It’s rarely true of any problem, and it’s almost never true of technology.

Australia’s under-16 social media law took effect on 10 December last year. I’ve said many times that the intent was wonderful. My view is that the application has been poor and the results worse, because the use hasn’t stopped, it’s gone out of sight.

Kids are clever. We raise them to see around corners, and then we’re surprised when they do.

This is PTFA in its most familiar form, Past Trauma and Future Anxiety, the way what has hurt us before shapes how hard we grab at what comes next. I’m of the generation making a lot of these calls, and I know it well.

Full of love, full of good intent, and quite sure the world is going to hell in a handbasket, which is more or less what our parents’ generation said about us.

The consequence for leaders runs well past social media. Ban a tool your people find useful and the use moves to personal phones and private accounts, where nobody can see what’s being shared or with whom. The rule written to remove a risk ends up removing your view of the risk instead.

A rule that keeps the behaviour where you can see it and shape it will nearly always do more good than one that drives it underground.

What I built, and why it asks first

After the 12:30 news, Phil asked about the other thing on my mind that day. A few hours earlier I’d soft-launched Choose Forward.

For thirty years I’ve kept a digital copy of everything I’ve done: every keynote, every conversation with clients, every article, the radio spots, many of them with Phil. More than one and a half million pieces of it now sit behind Choose Forward.

You type a business question at the top of the page, opening a second shop, whether a business has longevity, whether to hire, and you think it through with Mo, who is the digital equivalent of me. Don’t ask me where he got the name.

Mo isn’t working alone. There are seven characters in all, each carrying a different part of the work I do. One works the figures.

Reese does the research, a day ahead or twenty years out for your industry. Another looks at the human side, the people a decision touches. There’s also a way into the archive itself, so the old segments with Phil are in there too.

In April last year, on this same show, I told Phil that what we were seeing was the birth of the true digital twin, something that remembers what you’ve done and what you want, and can think through what to do next. Back in 2024 I’d written about digital selves that outlive us and sit in meetings on our behalf.

Choose Forward is a twin of a particular kind. It doesn’t stand in for me in a meeting. It sits beside you while you make up your own mind.

The design choice that matters most is how it starts. Before I was ever on a stage I did welfare and crisis work, on the streets and in prisons. Those are acute conversations.

There’s not much room, and you learn fast that the first question someone asks is often not the one they need answered. You have to find out what’s sitting underneath it.

I’ve taught Mo to inquire that way. Carefully, the way I’ve done it for thirty years, so that a chat about opening a second shop can surface the real question, which might be about cash, or a partner, or being tired. It asks before it answers.

I built it the way most people will build things from here on. I left the keyboard and talked to it. People call it vibe coding.

I fed it everything I’ve done and kept working through the issues with it out loud, until it understood what I was after even when I wasn’t saying it well.

HUMAND, with the decision handed back

This is HUMAND at its plainest.

It’s the three working together: the human, the machine and the AI. The AI underpins it all, doing the searching, the sorting and the remembering. The machine carries out the steps no person has time for.

The person keeps the judgement.

Most agents are built to finish the job. The ones worth trusting will be built to hand the important part back. Book the flights, yes.

Decide whether the trip is a good idea, that’s yours. An agent that knows where that line sits is worth more than one that’s faster.

The ripple effects

Voice stops being proof

For most of history, hearing someone was the end of the argument. It isn’t any more. Every family and every finance team will need a call-back rule: a payment asked for by voice gets confirmed on a number you already hold, or with a question only the real person could answer.

That’s a small cost for a household. For an organisation it’s a process change, and somebody has to own it.

Trust moves to whoever builds in the pause

The banks, insurers and service businesses that keep customers’ trust over the next few years won’t be the fastest. They’ll be the ones that slow a customer down at the right moment and put a real person on the line when it’s asked for. The hand-over becomes part of the product.

People will rehearse decisions with an agent before they ring a human

That changes the first meeting with an adviser, a manager or a mate. People will arrive having already thought it through, or having been talked into something. Across a whole customer base, the quality of the agent’s questions starts to shape the quality of people’s decisions, which means the questions themselves become something worth designing and auditing.

Small business may gain the most

A sole trader has no board, no strategy team and often nobody to think aloud with. I’ve heard the same sentence for years: I wish I had someone to think this through with. An agent built to think alongside, rather than decide for, goes straight at that gap.

The human matters more, in a narrower place

You can ask your best mate, and you should. Sometimes, though, you want to think without carrying anyone else’s feelings about it. I’m working on a general think-along app for life’s questions, and I expect it out in about four weeks.

It won’t replace the friend. It’ll be somewhere to get your thoughts straight before you talk to them.

What to do this week

Two small things.

First, agree a call-back rule with the people who could be impersonated to you: your kids, your parents, whoever approves payments at work. A known number, or a question only the real person could answer. It takes five minutes and it’s the cheapest protection there is.

Second, before your next big decision, ask yourself the question Mo asks. What’s the question underneath this one?

Choose forward.

If you’d like a short read every Tuesday on what’s coming and what it means for your work: welcome.morrismisel.com/join

---

Morris Misel is a foresight strategist, keynote speaker, and trusted adviser. He works with leaders, organisations, and boards to interpret what’s changing and make better decisions about what comes next. Heard by more than a million people a year on stage and in the media.

---

Questions people ask

What is an AI agent?

A chatbot answers your question. An agent goes off and does the steps for you: the searching, the comparing, the forms and the follow-ups, much like a travel agent who comes back with the flights, the hotel and the plan.

How do voice-clone scams work?

A voice or face is copied from short samples, then used in a message that asks for money or details in a hurry, often written to sound like a family member in trouble. The script is built to reach your heart before your head.

What is the best defence against an AI scam?

A pause. If a message asks for money or details in a hurry, the hurry is the warning. Ring the person on the number you already have, or the bank on a number you looked up yourself. The National Anti-Scam Centre's advice is Stop. Check. Protect.

How much did Australians lose to scams in 2025?

The National Anti-Scam Centre's Targeting Scams Report puts combined reported losses at $2.18 billion in 2025, up 7.8 per cent on 2024. Those are only the losses people reported.

What is a call-back rule?

An agreement with the people who could be impersonated to you that any request for money by voice or message gets confirmed on a number you already hold, or with a question only the real person could answer.

What is Choose Forward?

Choose Forward is a thinking partner built from thirty years of Morris Misel's work. You bring a business decision and think it through with Mo, his digital equivalent, who asks what you're really asking before he answers.

What does HUMAND mean?

It's Morris Misel's way of deciding who or what should do the work. The human, the machine and the AI work together, each doing the part it does best, and the person keeps the judgement.

Sources

This post is based on my segment with Phil Whelan on RTHK Radio 3 Morning Brew, 13 October 2026.

• National Anti-Scam Centre, Continued action critical to combat fraud as annual scam losses exceed $2 billion, 30 March 2026.
• National Anti-Scam Centre, Targeting Scams Report 2025.
• TheWrap, Netflix Recreates Gene Wilder's Voice With AI for 'Willy Wonka' Competition Series, 30 June 2026.
• eSafety Commissioner, Social media age restrictions, in effect from 10 December 2025.