The Autonomous Employee: When AI Starts Doing the Work

Simon Carver

[calm] Hey everyone, welcome back. I’m Simon Carver, here with Lachlan Reed, and today we’re talking about a shift that feels subtle right up until it doesn’t: the rise of the autonomous employee. Not an employee with better software. Not a smarter chatbot. We mean agentic AI systems that can take a goal and run with it.

Lachlan Reed

[chuckles] Yeah, g’day. This one matters because it’s not just, “Oh neat, the tool got faster.” It’s more like the tool’s wandered off, found the supplier, haggled the price, signed the paperwork, and sent you a summary before you’ve had your morning coffee. Bit of a wild one, really.

Simon Carver

Picture this. It’s 3:00 AM. No one’s online. But procurement contracts are being negotiated, supply chains rerouted, customer issues resolved. Work is happening, decisions are being made, and the workforce doing it never sleeps.

Lachlan Reed

[deadpan] So we’ve built the perfect employee... and accidentally made humans the bottleneck.

Simon Carver

That line lands because it’s uncomfortable. In a lot of workplaces, AI used to sit in the helper seat. Draft this email. Summarize that meeting. Suggest a few options. Useful, but still waiting for permission. Agentic AI is different. The human defines the outcome, and the system executes the chain of work end to end.

Lachlan Reed

Old model: you use software to do the task. New model: you tell the system what good looks like and it goes off and has a crack. So instead of, “Open the vendor list, compare quotes, send an email,” it becomes, “Find a supplier, negotiate, finalize the contract.” That’s not simple automation. That’s delegation to a non-human actor.

Simon Carver

And that distinction really matters. Automation traditionally shaved steps off repeatable work. Delegation hands over agency. The system is not just moving boxes down a line; it is interpreting goals, making choices, and acting across tools and processes.

Lachlan Reed

[excited] Which sounds efficient as all get-out. And it is, on paper. You get this kind of zero-latency economy vibe, where work keeps moving twenty-four seven, no lunch break, no waiting for Karen to come back from annual leave, no “let’s revisit this Monday.”

Simon Carver

But faster is not the same as wiser. That’s the trap. If the data is bad, the logic is flawed, or the system carries bias, then continuous execution doesn’t create intelligence. It creates continuous error.

Lachlan Reed

[laughs] Exactly. It’s like fitting a trail bike with a bigger engine when the front wheel’s loose. You’re not solving the problem, mate, you’re just getting to the crash site quicker.

Simon Carver

And this is where our series theme comes back in a fresh way. A lot of AI talk asks, “Will machines replace jobs?” I think the more precise question is, “What happens when the unit of work changes?” Because the worker is being redesigned, not just replaced. Human roles start shifting from execution toward orchestration.

Lachlan Reed

Which plenty of people haven’t fully noticed yet. They think they’re adopting a new tool. But under the bonnet, the organisation is quietly changing what counts as value. Less reward for doing the work. More reward for framing the goal, checking the result, and knowing when the machine’s gone sideways.

Simon Carver

So today we’re treating this as a practical guide. If AI is moving from assistant to operator, what changes for your role, your skills, your team, and your leverage? Because whether you’re a manager, founder, analyst, or someone early in your career, this is no longer abstract.

Lachlan Reed

[frustrated] Right, so let’s get into the messy bit. If agents are doing more of the doing, humans become supervisors of agents, designers of goals, and auditors of outcomes. Which sounds pretty tidy in a slide deck. In real life, it’s a bit less glamorous. You’re not necessarily in control. Sometimes you’re just the person left holding the paperwork when it blows up.

Simon Carver

Yes, and that’s why I’m cautious with the phrase “human in the loop.” It sounds reassuring, like a pilot calmly guiding the plane. In practice, it can mean the human is there mostly for liability containment. The system acts, the person approves, and accountability still lands on the person.

Lachlan Reed

And we didn’t eliminate work... we eliminated practice. That’s a dangerous trade. Junior staff used to learn by doing the boring bits: first drafts, basic analysis, small vendor checks, low-risk customer replies. Repetition built judgment. If AI absorbs those tasks instantly, where do people get their reps?

Simon Carver

That’s the vanishing apprenticeship problem. If the machine does the learning work, how do humans learn? You can’t really become excellent at outcomes if you never handled the pieces that taught you what good and bad look like. It’s like asking someone to direct a play when they’ve never been on stage, never built a set, never missed a cue.

Lachlan Reed

[skeptical] Yep. Folks say, “Don’t worry, people will move up the value chain.” Maybe. But move up from what? Careers used to be built in the middle layers of effort. We used to pay for effort. Now we pay for outcomes. And effort is where most careers lived.

Simon Carver

That shift also scrambles the skill hierarchy. The premium skills become critical thinking, domain expertise, orchestration, and clear prompting. Not prompting as a party trick, by the way. More like reducing ambiguity so systems can act safely. And domain expertise matters because someone still has to know what a good contract, a fair hiring choice, or a sensible lending decision actually looks like.

Lachlan Reed

So this isn’t just “learn the AI tools.” It’s learn to think in systems. If you’re managing multiple agents, you need to know what each one is doing, where the handoffs are, what could go pear-shaped, and when to stop the whole circus.

Simon Carver

Which brings us to governance, and honestly this might be the most important part. These agents can influence lending, hiring, vendor selection, all sorts of decisions with real consequences. The risks are familiar but sharper now: hallucination, bias, and reasoning that isn’t transparent enough for the people affected by it.

Lachlan Reed

[matter-of-fact] Yeah, we are delegating judgment to systems that cannot be held accountable. That’s the kicker. You can’t haul an algorithm into a performance review and ask why it favored one supplier over another. So agent governance stops being some niche compliance hobbyhorse. It becomes core business function.

Simon Carver

And the economic effects follow from that. Smaller teams can suddenly operate like much larger firms. Lean organisations can build digital departments instead of adding headcount. That’s powerful. But it can also hollow out mid-tier roles and widen inequality if the gains flow mainly to the people who own the systems.

Lachlan Reed

So productivity probably goes up. Output probably goes up. But the question is: who benefits? Because speed without governance doesn’t scale intelligence, it scales mistakes. And productivity without a pathway for people to grow can leave a lot of workers stranded on the side of the road.

Simon Carver

Which leaves the listener with a choice. You can be a passive user of AI, someone who gets a little faster for a while and slowly gets boxed in. Or you can become the person who directs intelligence: defines outcomes, audits results, and owns judgment where judgment must remain human.

Lachlan Reed

The future doesn’t belong to the smartest worker. It belongs to the one who can direct intelligence. And the ones who don’t learn that... won’t be competing with AI. They’ll be invisible to it. So have a think: are you doing tasks, or defining outcomes? Where could an agent replace what you do today? And where must a human still own the call?

Simon Carver

[warmly] That’s the work now. Thanks for being with us.

Lachlan Reed

Catch you next time, Simon.

Simon Carver

See you then, Lachlan. Bye everyone.