We’re handing entire workflows to AI agents before we’ve figured out how those workflows are supposed to work. What could possibly go wrong? A look at AI, startup chaos, and why we somehow automated ourselves right back to basic operations.
There are just so many quotable quotes here I almost don’t know where to start. OK, I will start with “We’re getting the order of operations wrong when we expect the AI to create clarity we haven’t created ourselves.” This is hands down the best Ops focused, original content on Substack. What a fantastic piece. Please keep these coming Katie!
This connects to something I see in hiring too: companies bolt AI onto broken onboarding and get a faster broken onboarding. The tools amplify whatever process already exists, good or bad, there's no neutral setting. What's your test for knowing a workflow is clean enough to deserve automation versus one that needs a human to redesign it first?
I’ve always thought Ops is underrated — yes strategy and product vision is cool but you still need to execute damn well. I’ve found the rigor to hand off to an AI intern so helpful in simplifying processes — humans are just too good at exceptions — this is so well written on what I’ve been thinking!
Oh, I love this!! 😂❤️ Building my own business and experimenting with AI, I'm seeing just how important that clarity is from the beginning. AI can absolutely amplify what we're doing but if we haven't defined what we're trying to accomplish or how things should work, what exactly are we amplifying?
Also, “you've just given your chaos a company email address” had me laughing. 😂
This is exactly where AI becomes an operating-system test.
If a workflow already depends on tribal knowledge, invisible handoffs, remembered exceptions and one person quietly filling the gaps, the agent does not magically create a system.
It exposes the fact that humans were compensating for one. The technology may be new, but the problem it reveals is basic: does the organization actually know how its work works?
What I find fascinating is the drive that leads us to decide that we have no other choice but to go faster and faster. We ignore that; giving the responsibility for performance to AI means we start losing our own position in the business machine.
A clear example is what I see happening to software engineers: they built agents that take care of everything, and now the amount of code generated with AI is such that the engineers who orchestrated it all have no capacity to review all that. How do they solve it? They build another agent to review the code.
I'm not saying we should stop using AI or setting up agents; what I'm saying is to pause and ask strategically how to use them. We need to think about our new role as well.
When you become a parent, people tell you that kids are really effective chaperones. Suddenly, you pay attention to your language, what you drink or eat in front of them, etc. That's what comes to mind when I think about automating a process using AI. :-) Suddenly it forces you to be very clear about what the steps are and to look closely at how many exceptions there really are!
There are just so many quotable quotes here I almost don’t know where to start. OK, I will start with “We’re getting the order of operations wrong when we expect the AI to create clarity we haven’t created ourselves.” This is hands down the best Ops focused, original content on Substack. What a fantastic piece. Please keep these coming Katie!
This connects to something I see in hiring too: companies bolt AI onto broken onboarding and get a faster broken onboarding. The tools amplify whatever process already exists, good or bad, there's no neutral setting. What's your test for knowing a workflow is clean enough to deserve automation versus one that needs a human to redesign it first?
I’ve always thought Ops is underrated — yes strategy and product vision is cool but you still need to execute damn well. I’ve found the rigor to hand off to an AI intern so helpful in simplifying processes — humans are just too good at exceptions — this is so well written on what I’ve been thinking!
An agent can't automate a decision nobody made. It just runs both versions faster.
Omg I felt all of this so deeply in my SOUL.
Oh, I love this!! 😂❤️ Building my own business and experimenting with AI, I'm seeing just how important that clarity is from the beginning. AI can absolutely amplify what we're doing but if we haven't defined what we're trying to accomplish or how things should work, what exactly are we amplifying?
Also, “you've just given your chaos a company email address” had me laughing. 😂
This is exactly where AI becomes an operating-system test.
If a workflow already depends on tribal knowledge, invisible handoffs, remembered exceptions and one person quietly filling the gaps, the agent does not magically create a system.
It exposes the fact that humans were compensating for one. The technology may be new, but the problem it reveals is basic: does the organization actually know how its work works?
Great article!
What I find fascinating is the drive that leads us to decide that we have no other choice but to go faster and faster. We ignore that; giving the responsibility for performance to AI means we start losing our own position in the business machine.
A clear example is what I see happening to software engineers: they built agents that take care of everything, and now the amount of code generated with AI is such that the engineers who orchestrated it all have no capacity to review all that. How do they solve it? They build another agent to review the code.
I'm not saying we should stop using AI or setting up agents; what I'm saying is to pause and ask strategically how to use them. We need to think about our new role as well.
When you become a parent, people tell you that kids are really effective chaperones. Suddenly, you pay attention to your language, what you drink or eat in front of them, etc. That's what comes to mind when I think about automating a process using AI. :-) Suddenly it forces you to be very clear about what the steps are and to look closely at how many exceptions there really are!
Nailed it, thanks for sharing!