Sponsored by
Hi, there!
Welcome to the 23rd edition of Work in Beta.
In this edition, we answer the question: when to use AI, when to leave it alone, and how to tell the difference for your own work.
Our free AI Starter Kits are the fastest way to go from "I should use AI more" to actually using it. Pick a tool, open the first lesson, and you'll have something working in 10 minutes. Start one here.
So let's dive in!
IF YOU ONLY HAVE 2 MINUTES
Image Credits: ChatGPT / Work in Beta
THE ‘HOW TO’ PLAYBOOK
Are You Using AI on the Wrong Work?
Every week, someone asks us the same thing: "When should I actually use AI and when should I leave it alone?"
They want a list. Good for this. Bad for that. We understand why: a clean list would be easy to follow. But we don't give them one, because the list doesn't exist.
Here's why. Take a financial reconciliation. If you know finance well, AI saves you an hour, and you'd catch any mistake it makes in seconds. If you don't know finance, the same task is dangerous. AI hands you a clean-looking answer, and you have no way to tell if it's wrong. Same task. Same tool. One person should use it. The other should be careful.
So the answer isn't a list of good and bad tasks. It's a way of looking at your own work. Let's start with one job - sales.
Let's Take One Job: Sales
A lot of a salesperson's work can go to AI: research on the companies they're about to pitch, the first draft of the pitch deck, a quick read on where they stand against their targets, the follow-up notes after a call. This is the work around the deal, and AI is good at it.
What stays with them is the judgment. Which accounts are worth chasing. Who to pitch to, and when. How hard to push, when to hold, when to walk away. The call with a buyer who's gone quiet. None of that goes to AI, that's where the real value of the job sits.
This isn't a guess. OpenAI's own sales team built an AI assistant that does exactly the first kind of work. It prepares meeting briefs, pulls account history, and writes call recaps. The average rep trades about 22 messages with it a week and saves close to a full day. But the rep still runs the meeting. The rep still makes the call. AI does the work around the selling. It doesn't do the selling.
That's the whole idea in one line: use AI to get more done on the work you already understand.
What Kind of Work Is It?
Look at any task on your plate. It falls into one of three kinds. And how much of your work lands in each one depends on how experienced you are.
Work you've done a hundred times. You know exactly what "good" looks like, so you can hand most of it to AI and fix what comes back in minutes. The more experience you have, the more work like this you own and the more you can safely pass off. Someone new has mastered fewer of these, so they have less to hand over. That's the real gap between a beginner and a veteran here. It's not who's better with the tool. It's how much they've already mastered.
Important work with a clear path. AI can build the first version, but you read it closely before it counts. This bucket grows with experience too. Take a client presentation. For someone who has built a hundred of them, AI writes the first draft and they shape it from there, no stress. For someone in their first month, that same presentation might be the only thing on their plate and they shouldn't hand off any of it. They don't yet know enough to tell if AI got it right, and building it themselves is how they learn. Same task. Different bucket. It depends on who's holding it.
Work that's all judgment. The negotiation. The pricing call. The hard conversation with an unhappy customer. Nobody hands this to AI, no matter how senior they are. This part stays yours.
So there is no fixed list of "AI tasks." As you get better at your job, more of your work slides into the buckets you can hand off. A beginner keeps more in their own hands on purpose. That's how the judgment gets built in the first place.
There's research behind this. Harvard and BCG found that when people used AI on problems they couldn't judge themselves, they did worse than people using no AI at all. The danger isn't using AI. It's using it where you can't catch its mistakes.
You're the Editor Now
There's one rule that sits on top of all three buckets. The moment you hand work to AI, your job changes. You stop being the creator. You become the editor. AI writes the draft. You decide if it's any good, fix what's wrong, and put your name on it.
How hard you edit depends on what's at stake. A quick internal summary gets a light pass. Anything going to a client, anything with money or a real decision attached, gets read line by line. But you never take what AI gives you and ship it untouched. The editor's seat is always yours. The more the work matters, the more editing it needs.
This holds for any role, not just sales. An operations person can let AI draft the weekly status update, but the call on which trade-off to make across teams stays human. A marketer can ask AI for ten versions of a headline, but the claim that has to be legally true gets checked by a person.
And here's the good news. AI isn't here to take your job. When Anthropic looked at how people actually use AI at work, most of it was people speeding up their own work, not machines doing the job instead of them. So the goal was never to replace yourself. It's to get the routine work off your plate so you can spend more time on the work only you can do, the work you're actually paid for.
The Mistakes We See People Make
Hunting for one universal list. People want a chart that tells everyone what AI is good and bad at. It doesn't exist. The answer depends on your work and how well you know it.
Handing AI work you can't check. If you don't understand the area, you can't tell when AI is wrong. And the better its answer sounds, the easier it is to trust by mistake. Polished and correct are not the same thing.
The beginner's trap. Taking AI's answer at face value and never building your own judgment. Early in a role, doing the work yourself is the point. Skip that, and you never learn to spot the errors.
The skeptic's blind spot. The experienced person who refuses to touch AI on work they could check in seconds isn't just losing time. The ground is moving under them and they can't feel it. The people around them are getting faster, the work itself is changing, and the gap only shows up once it's already wide. Don't be in that boat.
Final Thought
"When should I use AI?" was never going to have one answer for everyone. It comes down to two questions: what kind of work is this, and do I know it well enough to edit what AI gives me?
Run this week's tasks through those two questions and you'll know where AI belongs. Use it to do more of the work you already understand. Stay the editor. Keep the judgment that's yours.
WORK WITH US
Build With Us
Most professionals know AI can do more for them. The gap isn't awareness - it's knowing where to start, what to change, and how to make it stick.
That's what we work on through Work in Beta.
For individuals, we run working sessions - not teaching sessions, not agency engagements. You bring a real work problem and we figure out the AI solution together. You build, we guide. Think of it as borrowing our learning curve instead of building your own from scratch. You walk out with something working and the skills to keep going. When you get stuck, we're a message away.
For organisations, AI adoption is a people problem, not a technology problem. Your teams have the tools - what's missing is the translation layer between AI capability and daily work. Which processes to redesign, which habits to break, how to build genuine fluency - not just awareness. We help close that gap through hands-on training, process redesign, and deep adoption engagements. Not advisory, forward-deployed.
If any of this resonates, email us at [email protected] / [email protected] and we will figure out how to work together.