AI Lets You Spin More Plates. But Who Catches Them When You Stop?

"AI can make one person extraordinarily productive. That doesn't necessarily make the business any stronger."

I've spent most of my working life spinning plates.

Running businesses. Managing people. Juggling projects. Making decisions. Like most business leaders, there's always been more to do than time available.

Then AI came along.

Suddenly, I could spin more plates than ever. And the better I became at using it, the more I took on. It felt like extraordinary progress.

Until I started asking myself an uncomfortable question. Was I making the business more capable, or was I just becoming a bigger single point of failure?

The productivity trap

Think about what happens when you give a capable business leader more capacity.

They don't usually sit back and enjoy the extra time. They fill it. Another project. Another client. Another opportunity that's been sitting on the list for six months. That's not necessarily a bad thing. It's how businesses grow.

But AI takes this to another level.

Over the last couple of years, I've gone from individual conversations with AI to specialised assistants, multiple agents working on different tasks, and increasingly sophisticated systems where agents can check each other's work.

Initially, there was a lot of effort involved. Building context, checking outputs, questioning whether the time invested was actually producing a return.

Sometimes it felt like AI theatre. Impressive technology, elaborate systems, but was I actually getting anywhere faster?

Then the tools improved. My own capability improved. The systems became more reliable. And the productivity gains became very real.

I could move between complex projects faster. Have agents preparing work while I focused elsewhere. Get up to speed on something that had previously taken hours of reading.

But there was a catch.

The more I could do, the more I became responsible for. The hamster wheel hadn't disappeared. It was spinning faster. And much of what made it work still depended on me.

Your best AI user might be your biggest dependency

I've written before about what happens when a valuable employee leaves and takes their AI capability with them.

But there's another version of that problem. What happens when the person building and running these increasingly powerful systems is the business owner or a senior leader?

Think about the most capable AI user in your organisation. They might have developed their own assistants, refined instructions over months, built specialist workflows and figured out how to get consistently good results.

They're producing work at a level that would have been difficult to imagine a couple of years ago. But how much of that capability could someone else actually use?

The instructions might be buried in chat histories. Important context might exist only inside their personal AI projects. They know which outputs need checking, which agents work best and how the different pieces fit together. They've developed the judgement to make the system work. Someone else could have exactly the same AI subscriptions and still have no idea where to start.

You have an extraordinarily capable person. But you haven't necessarily built an extraordinarily capable business. And the more work that person takes on, the bigger the dependency becomes.

The shift from personal productivity to shared capability

This is one of the challenges we've been actively working on inside GenerationAI.

How do you take what a small number of highly skilled AI operators can do and make that capability accessible across an ordinary business?

Not by turning everybody into an AI expert. By making the work itself easy to pick up: what was done, why, and what’s still open, all kept somewhere the next person and their agent can find it.

We've been developing our internal systems with that objective in mind. And recently, we've started seeing the results as other people and their own AI agents begin working inside that environment.

The important breakthrough isn't simply that an agent can find a document.

It's that another person's agent can begin understanding the work, the decisions behind it and what needs to happen next, without somebody having to explain everything from scratch.

That changes the equation.

"Instead of the capability living entirely with the individual who developed it, you're starting to make it available to the next person".

The question shifts from how much one person can achieve to how much of that capability can be transferred.

The most important work might be what you leave behind

Here's something I've learned as we've developed these systems. When you finish working with an AI assistant, where does everything you've just accomplished go?

Traditionally, much of it stays in the conversation. You might save the finished document or send the email. But what about everything that happened along the way?

The decisions you made. The alternatives you considered. The problem you discovered. The reason you changed direction. The things you still haven't resolved.

All that context can disappear into a chat history that nobody else knows exists.

And increasingly, those conversations are disposable. You finish one, start another, move to a different agent or change tools. If the knowledge stays there, the next person starts from scratch.

So we've been developing a different discipline. At the end of a working session, the agent records what happened. What changed. What was decided. What's outstanding. What the next person needs to know.

That record goes back into the business's information environment, alongside the relevant files and instructions.

Not every conversation needs preserving. The important part is capturing enough of the working context for somebody else to continue. Then another person, using their own AI agent, can navigate to that work, understand the history and pick up where the previous person left off.

They'll still need judgement. They may need to ask questions or verify assumptions. But they're no longer starting with a blank page. And each time somebody contributes, the shared understanding can improve. That's when the work stops being something only one person knows how to do.

The foundations matter more as the capability grows

There's an important catch. If multiple agents are going to rely on the same information, that information needs to be trustworthy.

An outdated instruction, an incorrect decision record or a poorly managed permission can affect more than one person's work.

And as agents move from reading and drafting to changing information or taking actions, the potential consequences grow.

That's why the discipline matters.

Clear rules about what agents can access and change. Reliable records. Defined ownership. People who understand when to question or verify what an agent tells them.

You don't need to rebuild your entire information environment before getting started. But you do need to establish good practices from the beginning. Otherwise, instead of making good capability transferable, you risk making mistakes transferable.

Here's what I'd do on Monday

Find one person in your business who's become genuinely effective with AI. It might be your best employee. It might be you.

Pick one useful piece of work they regularly complete with AI assistance. Then try something. Ask them to finish a normal working session and leave behind everything another person would reasonably need to continue.

The relevant files. The instructions. The decisions made. The outstanding work. Any important assumptions or limitations.

Keep it in an approved business environment, with appropriate access. Then have somebody else pick it up using their own AI assistant.

And here's the important part. Don't let the original person explain everything. Watch what happens. Can the next person understand the work? Can their agent find the right information? Do the decisions make sense? Can they identify what's outstanding? Where do they get stuck?

Every point where they need to go back to the original person tells you something. It shows you where the capability still depends on an individual rather than the business.

You don't need to fix everything. Start with one workflow. Close the gaps. Try again. If another person can reliably pick up that work, you've achieved something much more valuable than making one employee faster. You've made a piece of your business less dependent on the person who built it.

A business that keeps getting stronger without you

We're entering a period where one capable person can do things that previously required several people. That's a remarkable opportunity. But I don't think the businesses that benefit most will necessarily be the ones with the most impressive individual AI users.

They'll be the ones that figure out how to make that capability transferable.

Where a new person doesn't have to spend months reconstructing how everything works. Where decisions and lessons don't disappear into forgotten conversations. Where one person's improvements become something the next person can build on. And where the business can keep getting stronger without becoming increasingly dependent on a handful of extraordinary individuals.

Because spinning more plates is impressive. Building a business that doesn't depend on you keeping them all spinning is something else entirely.

The real AI breakthrough isn't how much more one person can do. It's how much of that capability the business can keep.

“The information you have spent years creating is about to become a real competitive advantage, if AI can actually reach it. Start getting it ready.”

Find out how much of your AI capability depends on one person

Most businesses have someone who’s become brilliant with AI. Few know how much of that capability the business could actually keep if that person stepped away, or where it’s already creating risk.

The 5 Minute AI Risk Snapshot shows you where your AI use is building real business capability, and where it’s sitting with one individual. You’ll get a personalised report with practical next steps.

Take the 5-Minute AI Risk Snapshot

--

About the Author 

Mark Lucas is Co-founder and Managing Partner at Generation AI, where we help New Zealand business leaders build real AI capability.

Information provided is general in nature and does not constitute legal advice