AI was supposed to make my job lighter. But as the easy stuff disappears, what’s left is grueling, cognitively demanding work. I’m faster, sure. But I’m also exhausted. And it’s taught me something I think a lot of leaders are about to learn the hard way: when AI handles the easy stuff, what’s left is a job you have to design for yourself.
When AI does the easy stuff, what’s left is the hard stuff.
Here’s something I didn’t see coming.
I’m a person who took 18 months to realize my laptop was a touchscreen. And somehow, over the last year—while running a 25-person strategy execution consulting firm—I’ve become a low-key AI geek. I build constantly. I tweak incessantly. I track the time savings on a spreadsheet because I’m a consultant, and that’s what we do.
By my honest count, I’ve automated, semi-automated, or significantly sped up 25–35% of my job. And contrary to everything you read about AI slop, the quality of my work has gone up. Not down. Only because I trained the AI to be mean to me. LOL.
My meeting load hasn’t changed. But the work that used to fill the gaps between meetings — drafting, summarizing, formatting, first-passing — gets done in a fraction of the time. I should feel lighter.
I don’t. I feel more tired.
The more I talk to other people who’ve put in the reps with these tools, the more I hear the same thing.
I’ve never had more time back, and I’ve never been more wiped out at the end of the day.
That’s a paradox worth unpacking. Because what’s happening is not what most people think.
AI takes as much as it gives.
Companies laying off people in anticipation of AI have clearly not used AI very much. Sure, it will eventually be able to do commoditized, explainable work at a faster pace than humans. But in my two years of heavy use and experimentation, I can say two things with some confidence:
- Removing routine work shortens, eliminates, or automates tasks. That doesn’t mean it can replicate, replace, or outperform jobs.
- When most of the easy stuff is done by a robot, you realize how much hard stuff was held hostage by the easy stuff. Your “someday I’ll” list very rapidly becomes a “tomorrow I’ll” list.
And there’s a third thing.
Heavy AI users are discovering that easy work was doing more for them than they realized. The drafting and formatting and first-pass summarizing—that wasn’t just filling time. It created cognitive pacing—lighter stretches between the hard stuff that let the brain recover between demanding tasks. And it handed you a built-in to-do list so you didn’t have to decide what mattered most on any given morning.
When that work disappears, what’s left is a job made almost entirely of the cognitively demanding stuff — with no defaults for how to structure it or recover from it.
When AI handles the easy stuff, what’s left is a job you have to design for yourself.
If you lead a PMO or a portfolio function, your team is going to hit this wall too, if they haven’t already. The productivity gains feel like pure upside right until they don’t.
Over the last year, I’ve had to learn an entirely new way to do my job as I share it with AI. It hasn’t been easy. But in case you’re feeling what I’ve been feeling, here are six things I’ve learned about leading in a post-AI work environment — and why I think PMO leaders need to be thinking about them now.
1. The hard work has become identifying the hard work.
Each year, and well before AI entered the scene, I run my work through a quadrant — what I call a Strategic Value Audit: Love it to Hate it, and Uniquely Good at to Passable/Actively Bad At. The sweet spot is work I love and I’m uniquely good at, especially where it overlaps with what’s strategically important for the organization I lead.
That quadrant used to be a nice annual exercise. Now it’s essential. When AI clears the routine work off your plate, there’s no default list of tasks anymore. The list is yours to write.
Without as much routine work crowding my week, I have to make constant decisions about what is strategically important—the “wicked problems” worth solving. Strategy isn’t an annual or quarterly event anymore. It’s how I prioritize my week and month.
A question I find myself asking a lot: “What will future Sara wish current Sara had started working on now?”
One example of a problem I’m tackling now: how do you let people experiment with AI fast enough to learn, without letting them make architectural or security decisions that the whole enterprise will have to unwind later?
That’s the kind of question I used to want to work on, feel like I should be working on, and almost never block two days to think through. Now it might be 15 hours of my week. And it’s well worth the investment.
If you lead a PMO or a portfolio function, you have a version of this question. Maybe it’s how AI changes your delivery methodology. Maybe it’s what your PM role looks like in two years. The point is, these questions don’t announce themselves. And the routine work that AI just took off your plate was the thing preventing you from ever getting to them.
Once I identify the hard work worth doing, I still have to evaluate it against four things:
- Likelihood to advance the strategy
- Speed-to-payoff
- Consequences of delay
- How much energy it burns over the course of a day
That last one matters more than I expected. Deep strategic thinking degrades as the day wears on. The question at 3 pm is not the same question at 7 am.
2. What you figure out in one area has to be applied to five others.
A question I can’t stop thinking about: Can you teach a PM fundamentals class without at least 50% of it being AI-enablement? Prompt libraries to help, skill files to speed up the work?
If the answer is “no” — and I think it is — the implications don’t stop at one class. We need to rebuild our content for the world we live in. Equip our own PMs with the same skills. Talk to PMO Directors about how to do this with their teams. Weave it into our newsletters. And we need to make the philosophical shift internally as quickly as possible.
One question. Five places it needs to go. And that’s just the version I can see from where I sit.
Before AI freed up my thinking time, I didn’t generate these kinds of cascade questions as often.
Now they come up constantly. And the work of tracking where a single insight needs to show up across the organization is real work.
If you run a PMO, think about what happens when someone on your team figures out a better way to build a dashboard, codes a tool to run better stand-ups, or uncovers a missed requirement that touches five other projects and could have been caught with a better process. That insight doesn’t belong to one project. It belongs to the methodology, the training, the governance model, and probably three other places. Somebody has to see that web and work it. In the post-AI job, that somebody is you.
3. Energy management is the new time management.
I now wake up at 5. I get three hours of deep work before most people start their day, and if I can, I stay clear all the way until 11. That’s my deep work window. Meetings ideally happen after that, because by then my strategic thinking capacity is shot.
Before AI took the routine off my plate, there was enough lightweight work mixed into the day to create natural recovery. Those tasks gave my brain a break between the hard stuff without me having to think about it.
Now the lightweight work is done in minutes, and what remains is almost all high-effort. I have to design around that. For me, that looks like doing the deep work in an uninterrupted block while my mind is fresh…and waking up early to do it.
The meetings, the collaboration, the reviews of other people’s thinking and output…that all happens when my energy needs a mode switch.
And even with all of that finagling, energy has become the bottleneck resource.
For PMO leaders, this is worth paying attention to—especially if your team is starting to use AI for status reports and risk logs. The highest-performing PMs who get faster at routine work aren’t going to coast on the extra time. They’re going to fill it with harder work. And if nobody helps them think about how to pace that, you’ll get the AI productivity paradox: more capacity, more exhaustion.
4. Output takes longer now, and you have to stop apologizing for it.
I struggle with this one. By nature, I’m a fast mover. I want quick decisions we can learn from faster. I want something tangible that communicates thoughts and ideas. I want items to check off my to-do list.
But hard work takes longer to show a result. Choices you make now have knock-on effects. And even when you make a decision, you have to triage where you apply it and how fast. Updating one training class is quick. Adjusting our philosophy, staying ahead of AI capabilities, building resources, and scaling all of that across a program takes months — and has to get weighed against other hard work I could be doing.
I can spend an entire week working on a strategic question and have nothing tangible to show for it except a clearer understanding of the problem and a plan for how to collaborate with the right people on a solution. A year ago, that would have felt like a wasted week. I’m starting to recognize it as the week the real work happened.
If your PMs are starting to use AI and suddenly have more capacity for harder work, they’re going to run into this too. The deliverable-per-week cadence that made sense when 40% of the job was formatting and first-drafting no longer makes sense when the work shifts upstream. Leaders who still grade PMs on busyness and artifacts will misread what’s happening. The better measures are the ones that were always the better measures: value delivered, crises averted, lessons applied across the portfolio. The difference is that now there’s time to focus on them.
5. AI makes it easy to think alone. That’s not always a good thing.
I’m an introvert. I do my best thinking in the quiet. Brainstorming in a room is overstimulating to me, and I tend to dominate the conversation in unhelpful ways simply because I’m trying to follow it.
AI has made this worse because it simulates working with someone at a pace I’m comfortable with. I can think through a brand idea or a new concept for a newsletter, ask the AI questions, iterate on the answer, and feel like I’ve had a productive working session. Sometimes that’s true. But sometimes what I actually needed was a fellow expert or a motivated thinker — and I skipped that step because the AI interaction felt like enough.
How far do I go solo before talking to a person? If I’m developing a content idea, when should I bring in the SEO person to pressure-test whether it’s of interest to our ideal customer or just to me? If I’m thinking about integrating AI into our classes, when should I bring it to our content team? Too soon and it’s overwhelming. Too late, and it feels like I’ve already decided, and quite possibly decided wrong. Too solo, and I might get buried in complexity when I have team members who are really good at carve-outs.
This one doesn’t have a clean answer yet. But the pattern is worth naming:
AI can make solo thinking feel collaborative when it isn’t.
And for anyone whose natural preference is to think alone first and talk to people later, that’s a trap worth watching for.
If you lead a team that’s starting to use AI as a thinking partner, pay attention to when people stop looping each other in. The bigger risk: they’ll use AI well enough to skip the human conversations where the real pressure-testing happens.
6. You need a heuristic for when to stop building and start thinking.
This is a tough one. Automation opportunity is vast. And just when you think you’ve hit the ceiling of what you can get assistance with, new models or features come out that let you do more. Building has become a skill that needs practice. Not everything you build will save you time. But it takes time building to get a sense of what’s worth building and what isn’t. It’s a catch-22 that’s easy to get lost in.
I’ve fallen into it more than once. I’ll spend a morning building a workflow that saves me 20 minutes a week, and realize I just burned three hours of deep-work time I can’t get back. Or I’ll resist automating something for weeks, finally build it, and wonder why I waited so long. The decision itself — build or just do the work? — creates its own kind of fatigue, especially when the opportunity comes up multiple times a day.
For now, I’ve landed on a set of questions I ask myself before I start building. They’re not a formal framework. They’re more like the things I’ve learned to pause on after enough wrong calls:
- How often do I do this, and does it have to be now? Some things are better automated the next time they come up, on a lighter day.
- Is there a thinking benefit to doing it myself? Some tasks teach me something in the doing.
- Is there a risk to not doing it myself? Client work, financials, anything where a mistake is expensive.
- How long will it take to set it up right? If setup takes longer than just doing the work, it’s not worth it yet.
- Will I learn something or codify something useful down the line? Sometimes the build is the learning, even if the time savings are marginal.
- How likely am I to rebuild this when the process or models change? A tool I’ll have to rebuild in three months has a different payback than one that’ll last a year.
- Does it fit the pattern of builds that have worked? I look at my track record. The builds that saved me time tend to have things in common.
Having this list hasn’t eliminated the tension. But it’s shortened the internal debate from 20 minutes to 2, which, over the course of a week, adds up to a lot of reclaimed thinking time.
We all just became the business analysts of our own jobs.
If you read back through those six things, there’s a thread connecting them all. Identifying the hard work. Figuring out where insights need to go. Designing around your energy. Accepting that output takes longer. Knowing when to stop building and start thinking. Knowing when to stop thinking alone.
None of that was in my job description a year ago. But all of it is work I now do constantly—defining requirements, sequencing priorities, deciding what to automate and what to protect, mapping where a single decision ripples across the organization. That’s business analysis work—applied to my own job, in real time, while also doing the job. It’s what I’d call the post-AI leadership redesign: the ongoing work of figuring out what your job even is now that AI handles the parts you used to default to.
What I didn’t expect: it’s becoming less mechanical and more intuitive. The first few months of this felt like building the plane while flying it. Every decision was new. Every tradeoff took energy. Now some of it is starting to feel like instinct — I know which problems are worth 15 hours of my week and which ones aren’t. I know when I’ve been solo too long. I know when I’m building to avoid thinking.
That shift from science to art is worth noting because it means this gets easier. Not easy. But easier. The six things I described aren’t a permanent tax on your energy. They’re a learning curve — and like most learning curves, the steepest part is the beginning.
If You Only Do One Thing
I’m going to do something I don’t normally do and point you to someone else’s work. Nate B. Jones, who writes about AI and productivity at his Substack newsletter, is one of the best thinkers on AI that I read, and he wrote a post recently that I love. In it, he walks through how you can do an honest audit of your work. The key insight for me: Using the time savings from commoditized work to do more commoditized work, just faster, is the wrong move. Worth a read this week.