Stop Automating the Mess
I've been spending a lot of time lately thinking about how we use AI inside our delivery organization.
And I'll admit, my instinct is usually the obvious one:
How can we automate this?
Recently, I was looking at one of our project management processes.
We had a RAID log tracking risks, actions, issues, and decisions. We had delivery data sitting in another system. We had meeting notes. Then every week, someone was taking all of that information, reconciling it, summarizing it, and manually building a status report.
Think about that for a second.
We already had the information.
We were paying talented people to move it around.
So naturally, my brain went to AI.
Let's automate the status report; and we did.
It worked really well.
But the more I looked at it, the more I realized that automating the report wasn't actually the most interesting part.
The better question was:
Why did this require so much manual work in the first place?
That's a question I've started asking a lot more lately.
How Did We Get Here?
Most organizations have processes nobody really designed.
They evolved.
Someone needed a report, so we created a spreadsheet.
Someone else needed visibility, so we added a dashboard.
Leadership needed an update, so we created a PowerPoint.
The system couldn't quite do something, so someone created a workaround.
Then that person left.
Someone else inherited it.
Three years later, it's called "the process."
I've seen it throughout my career.
And I'm sure I've created a few of these myself.
That's organizational muscle memory.
We don't necessarily know why we're doing something anymore. We just know that every Thursday, somebody does it.
Now AI comes along and gives us the ability to do that thing significantly faster. That's valuable; but there's a trap in there too.
Because if we're not careful, we're going to spend the next few years becoming incredibly efficient at doing things that never needed to be done in the first place.
AI Can Make a Bad Process Faster
There's a lot of conversation right now about productivity.
Hours saved.
Tasks automated.
Content generated.
Those things matter.
But imagine someone spends four hours every Friday pulling information from three systems, putting it into Excel, summarizing it, and creating a PowerPoint.
AI gets that down to 20 minutes.
Awesome. We saved almost four hours.
But I'm increasingly interested in what comes next.
Why were we doing all of that?
Maybe two systems are capturing the same information.
Maybe the spreadsheet exists because the systems don't talk to each other.
Maybe half the PowerPoint exists because that's simply how we've always presented the information.
Maybe the executive reading it really only needs three things:
What's happening?
What are we worried about?
What decision needs to be made?
Now we're having a much more interesting conversation.
We're not automating a task anymore.
We're redesigning the work.
Sometimes the Workaround Is You
This one has been a little more personal for me. As I've moved into larger leadership roles, I've had to recognize how easy it is to become the system yourself.
A question comes up?
Ask me.
Need someone interviewed?
I'll do it.
Delivery issue?
Pull me in.
Need context from three different conversations?
I probably have it.
Something isn't moving?
I'll jump in and push it.
There's a certain satisfaction that comes from being that person.
You're useful. People trust you. Things move. Problems get solved.
And for a while, that can even look like really effective leadership.
But it doesn't scale.
I've been learning that one of the hardest transitions in leadership is recognizing that being capable of solving something doesn't necessarily mean you should be the person solving it.
Sometimes the better leadership move is to slow down long enough to ask:
Why did this need me in the first place?
Was ownership unclear?
Was someone missing context?
Was the process broken?
Did we never give someone the authority to make the decision?
Do we need another person?
Or have I simply trained the organization to come to me?
That last one can sting a little.
Because if something only works because one person knows where everything is, who to call, and what needs to happen next...
You don't have a process. You have a person.
I've been that person plenty of times. I'm trying to be that person less.
Not because I want to be less involved; but because I want to build teams and systems that don't require heroics to work.
Start With the Problem, Not the AI
That's changing the order of the questions I'm asking.
Instead of starting with:
How can AI automate this?
I'm trying to start with:
Why does this exist?
Then:
What outcome are we actually trying to create?
Where is the friction?
What should a person own?
What should the system own?
Where do we actually need judgment?
And only then:
Where can AI help?
Sometimes AI is the answer.
Sometimes it's better integration.
Sometimes it's a clearer process.
Sometimes someone just needs ownership.
Sometimes it's eliminating a meeting.
And occasionally, the answer is my personal favorite:
Stop doing it.
Give the Time Back to People
There's another piece of this that I think matters.
When we save someone four hours with AI, what are we actually doing with those four hours? Because if our answer is simply, "Great, now you can produce four more hours of stuff," I'm not sure we've figured this out.
One of my favorite movies is The Dark Knight Rises, and there's a Bane line that always makes me laugh in this context:
"Give it back to the people."
Now, I'm pretty sure Bane wasn't talking about AI productivity or organizational design, but stay with me.
What if that's actually the goal?
Give the time back to the people.
I'd rather have a project manager spend that time talking to their team.
I'd rather have a consultant spend it understanding the customer's actual problem.
I'd rather have a leader spend it coaching someone.
I'd rather spend it looking across the organization and asking why the same problem has appeared three times instead of personally jumping in to solve it a fourth.
And frankly, maybe sometimes we just give people some of their time back.
Not every efficiency gain has to immediately become another unit of output. Maybe somebody gets to think. Maybe they get to learn something. Maybe they log off at a reasonable time.
That matters too.
AI shouldn't just help us do more work.
It should help us spend more of our time doing the work that actually matters.
So yes, in the least Bane-like way possible:
Give it back to the people.
The Leadership Part of This
The more I work with AI, the more convinced I am that the interesting part isn't really the AI. It's what forces us to reconsider work.
And for leaders, I think that means being willing to challenge some things we've historically rewarded.
Being the hero.
Knowing everything.
Being in every meeting.
Making every decision.
Being the person everyone calls when something goes sideways.
I've built some of my career by being good at those things.
But I'm increasingly convinced the next version of leadership looks different.
Build the system.
Create clarity.
Develop people.
Give them authority.
Remove friction.
Use technology intelligently.
Then get out of the way enough for people to surprise you.
That's harder than jumping in and fixing something yourself.
But it's also how you build something that lasts beyond you.
Find Your Mess
In my last piece, I wrote about Defining the Gray - the intersection between people, systems, and tools. I'm realizing that a lot of the mess lives there too.
It's in the handoffs, the spreadsheets between systems, the meetings between teams, the reports built from other reports, and the person everyone goes to because they know how everything actually works.
That's where I'd challenge you to look.
Find one process in your organization that everyone accepts but nobody particularly likes.
Don't start by putting AI on top of it.
Pull it apart.
Ask why every step exists.
Ask what happens if you remove one.
Ask whether the system is supporting the person or whether the person is compensating for the system.
Then figure out where AI belongs.
You might automate it.
You might redesign it.
You might give it to someone else.
You might kill it entirely.
Just don't automate the mess.
And if you've got one of those messy processes, the weird handoff, the report nobody really understands, the "we've always done it this way" problem; send it my way.
I genuinely enjoy pulling these things apart.
Because increasingly, I think that's where some of the best conversations about AI, systems, and leadership start.
Not with:
"What can AI do?"
But with:
"Why the hell are we doing this in the first place?"
Comments