I keep hearing some version of the same question from leaders and teams:
Why, if we have all of these AI tools now, does work not actually feel easier?
And from what I’m seeing, I think part of the problem is that we’re treating AI as though its existence should automatically make work easier.
But AI is still a technology.
It needs to be calibrated. It needs to be adapted. It needs to be applied to the right problems.
And, maybe most importantly, we still have to identify the return on investment or business value we expect it to create.
Otherwise, we’re just adding another tool to the tech stack.
I think we forget that sometimes because AI feels like another person.
But it isn’t one.
AI doesn’t have all the context you think it does
Recently, I watched this play out in a real workflow.
A team was using AI to help build out feature requests from customers and create more specificity and granularity as those requests moved from product to engineering.
In one case, the work had already been intentionally scoped down. What remained should have been a relatively simple engineering task.
AI expanded it back out.
It interpreted the request without the context behind the previous decisions and turned it back into a larger engineering and product effort.
The result?
Two additional meetings and four additional days of work evaluating something the team had essentially already decided should not be included.
The tool that was supposed to make the process more efficient actually created more work because there was now an extra participant in the process that didn’t have the context.
And that participant was AI.
AI is turning more of the workforce into managers
This is one of the changes I don’t think we’re talking about enough.
AI is effectively turning a much larger portion of the workforce into managers.
Not necessarily managers of people. But managers of work.
Suddenly, people who never spent much of their time mastering the art of delegation are being asked to delegate work to AI.
And they’re not delegating to a senior employee who has years of organizational context and can interpret what they probably meant.
They’re delegating to something much closer to a junior employee that needs explicit direction.
What are you asking it to accomplish?
What context does it need?
What does “done” actually look like?
What shouldn’t it do?
And once the work comes back, how are you evaluating whether it’s actually right?
Those are management skills.
Historically, we haven’t focused on training those skills to the entire workforce. But now we’re handing everyone AI tools and expecting them to know how to do this.
We communicate with AI as though it already knows what’s in our heads
There’s another piece of this that I think is easy to underestimate.
Humans carry an enormous amount of context around with us.
There’s a concept in The Culture Map around high-context and low-context communication. In a low-context situation, we assume the other person doesn’t know what we’re talking about, so we naturally provide more information. In a high-context situation, we assume more shared understanding.
I think a lot of us interact with AI in a surprisingly high-context way.
It feels conversational. It responds intelligently. Sometimes it even feels a little bit like talking to your own brain.
So we forget how much information is still sitting only in our heads or the connections we naturally made that aren't obvious.
We give it the request without the history.
We give it the task without the decisions that led to the task.
We give it the output we want without explaining why that output matters.
And then we’re surprised when it doesn’t interpret the situation the same way we do.
This is probably why so much of what we’re learning about prompt engineering eventually comes back to context.
The problem isn’t always getting AI to generate something.
It’s getting what’s in your head into a form AI can actually work with.
Faster isn’t a return on investment
There’s another problem underneath all of this.
Organizations are implementing AI because they want to go faster.
But making something faster is not, by itself, an ROI.
Faster at doing nothing valuable is still doing nothing valuable.
And now AI makes it possible to do a lot more of it.
That’s why I think organizations need to go back to a much more basic question:
What actually has to get done?
What can this business not live without accomplishing today?
What is the 20% of our work that’s driving the majority of the value?
What work is taking significant time and moving the business forward?
Start there.
Instead, I often see organizations start with:
“What can AI do?”
That’s a very different question.
Meeting notes are a perfect example
AI-generated meeting notes have become almost ubiquitous.
And on the surface, this makes perfect sense. Meeting notes take time. AI can generate them. Great.
But what are the meeting notes actually supposed to accomplish?
Are you generating them so a customer receives a clear record of what was decided?
Are they supposed to create alignment between teams?
Are they documenting ownership?
How quickly do they need to be delivered?
Who needs to review them?
What needs to happen because those notes exist?
Or are we generating meeting notes because AI can generate meeting notes?
I’ve also started seeing the inverse problem:
“AI is taking notes, so why do I need to pay attention in this meeting?”
Now the efficiency tool may actually be creating a bigger problem.
We have the notes.
But we have less shared understanding.
My personal filter for AI is pretty simple
I’ve started using a few basic rules when I think about where AI belongs in my own work.
If this wasn’t on my list of things that absolutely needed to get done today before AI, it shouldn’t suddenly be on my list because AI can help me do it.
If it’s something I would have delegated anyway, it shouldn’t necessarily become mine again just because I can delegate it to AI.
If I get to the end of the week and won’t notice whether it happened, I probably didn’t need to do it.
And if I can’t see how doing it creates some kind of business return, I have to seriously question why I’m spending time automating it in the first place.
AI should help us spend less time on work that doesn’t matter.
I’m worried we’re sometimes using it to make ourselves capable of doing more of that work.
The real problem may be leadership clarity
This is where I think the conversation gets uncomfortable.
Sometimes AI isn’t making work easier because the organization hasn’t actually decided where it’s going.
I’ve worked with leaders who are hesitant to set that direction because they’re afraid to be wrong.
And I’ve seen organizations where the direction is essentially:
“We need to make X amount of money.”
That’s a financial goal.
It’s not necessarily a leadership direction.
Compare that with one of the clearest operating environments I’ve experienced.
A leader implemented EOS—the Entrepreneurial Operating System—and the company identified three things that absolutely had to happen in the next 90 days.
The message was essentially:
If nothing else gets done in the next 90 days, these three things have to happen.
Then we evaluated our work against them.
Does what I’m doing get us closer to one of those three things?
Or is it just busy work?
You would hear the leader ask it in meetings:
“How does that get us to our goal?”
Over time, everyone started thinking that way.
And interestingly, narrowing the focus didn’t mean we accomplished less.
We got those priorities done faster, and the company ultimately created more traction in those 90 days than I’d seen before.
AI could make an organization operating that way extraordinarily fast.
But without that clarity, AI can just help everyone produce more things.
One of the biggest red flags I hear now is “we can just…”
“We can just do this quickly with AI.”
“We can build that really fast.”
“We can just vibe code it.”
Sometimes that’s completely true.
These are incredibly valuable tools.
But “we can make it quickly” and “we should make it” are two different decisions.
Before asking how quickly AI can do something, I want organizations to ask why they’re doing it in the first place.
Because when that conversation disappears, that’s often where I start seeing misalignment.
Teams are incredibly busy. They’re producing more than ever.
But people start asking:
What does my job actually turn into?
What’s the goal?
What value do I really provide?
Those aren’t questions I think we should dismiss as fear of AI.
They’re important organizational questions.
Start with what the business needs. Then ask what humans uniquely bring to it.
When someone starts questioning their value in an AI-enabled workplace, I don’t think the answer is to compete with AI.
I try to bring the conversation back to individual strengths.
What can you uniquely do that a tool or technology can’t replace, but can enhance?
I’m probably a little contrarian here because I’m not a huge believer in spending the majority of our time strengthening every weakness.
I would rather identify someone’s strengths and double down on them.
Then build systems around their weaknesses so those weaknesses don’t negatively impact the work.
And I think the same logic applies to AI.
Start with:
What is the business trying to accomplish?
Then:
What can you uniquely do to help it get there?
And only then:
What tools do you have at your disposal that can help you do that faster or better?
When we flip the order into those places, that’s when I see AI become incredibly powerful.
The goal isn’t to find more things for AI to do.
The goal is to know what actually matters—and then use AI to help us get there.