Some companies are now spending more on AI than they were spending on the people the AI was supposed to replace. And I think that's an us problem, not an AI problem.
I read an article about a company that ran up a $500 million bill on AI usage without noticing. Five hundred million.
But it isn't an tool problem; it's an amplification of a much larger problem we've had for a long time that we just got away with before.
The real issues are a lack of:
- Clear priorities
- Shared vision
- Consistent planning
- Reporting loops
- Systems for what gets done and what doesn't
- The right kind of leadership oversight
- Etc.
These are messes we've always had. The problem is, when you put AI on top of a messy system, you just get a really big mess.
So gaps in your team, direction, leadership, vision, etc. can and will get a HUGE magnifying glass on them when you task AI with scaling everything out.
Before, you could get away with a little chaos because human bandwidth was the natural ceiling.
But now that ceiling is getting removed, and the chaos has nowhere to stop.
People are quick to laugh and say, "see, AI is stupid." But that's a shallow interpretation of what's really happening (and this WILL keep happening).
The better thing to focus on is…how do we get much more aligned on what we're doing, why we're doing it, and how?
There's a concept in software development called spec-driven development.
Before you write code or generate outputs, you define the behavior you want, the constraints around it, the inputs it should accept, the outputs it should produce, and the criteria for what "done" looks like.
In other words, it forces you to get clear on your what, why, and how before you go kamikaze on building things.
This concept is essential with AI development, but it can and should be applied to our work at large.
Do people know what they're building, why they're building it, who it's for, and what "done" looks like before they start?
We've been running on the assumption that smart people will figure it out as they go, but that doesn't work as well now.
The same fuzziness that used to cost a few hours of rework now costs us hundreds of hours, or in some cases hundreds of millions of dollars.
The same discipline that makes AI development work is the same discipline that makes our teams work. The concept of direction, leadership, and clarity aren't new.
Now more than ever, we need to adopt a spec-driven approach to work.
- What exactly are you trying to accomplish with [fill in the blank]?
- Who is this actually for? Has anyone asked for this or are you just assuming they want it?
- What does success look like? Or could this accidentally go on indefinitely without anyone noticing?
- What are you intentionally not doing? And are you anticipating scope creep before it happens?
- What boundaries are in place to keep this from going too far?
Then let AI help inside those guardrails instead of getting lost in possibilities.
AI is a force multiplier, and a multiplier doesn't care what it's multiplying. You can multiply order or you can multiply chaos.
And at the rate our capabilities are growing, it's extremely important that we get things in order before we let AI make the chaos really expensive and hard to fix.