The Cost of Thinking: How AI Is Changing Product Management
Yesterday I bought two pieces of Parmigiano Reggiano.
5€ each.
Total: 10€.
Coincidentally, that was exactly the same amount I had set as my daily OpenAI API budget for experimentation.
The cheese will probably survive a few weeks.
The AI budget?
Usually gone by the afternoon.
And that’s when I noticed something interesting.
When I eat the cheese, I don’t think about the price.
When I experiment with AI, I constantly look at the credit balance.
How much did that prompt cost?” “Should I run this again?” “Is this worth it?
It’s a strange shift.
For the first time in decades of software development, we’re becoming aware of something new:
The cost of thinking.
Software Was Always About the Cost of Building
For years, the industry optimized one thing: building software cheaper and faster.
We invested massive effort into improving:
-
developer productivity
-
CI/CD pipelines
-
infrastructure cost
-
deployment speed
-
cloud efficiency
All of that focused on the cost of execution.
Once the idea existed, the system was designed to build it as efficiently as possible.
But product work never really starts there.
Before building anything, product teams spend enormous time on thinking work:
-
exploring ideas
-
interpreting user feedback
-
connecting signals across teams
-
shaping product strategy
-
deciding what not to build
Historically, this thinking work was slow but free.
Now it’s fast - but metered.
AI Turns Thinking Into a Resource
Large language models changed something subtle but fundamental.
They make thinking tasks dramatically faster:
-
summarizing feedback in seconds
-
generating solution approaches instantly
-
analyzing patterns across user signals
-
drafting product documents or strategies
But every interaction has a cost per token.
Suddenly product teams experience something new:
Thinking is no longer unlimited.
It’s measurable.
And once something becomes measurable, we start optimizing it.
The Real Bottleneck in Product Teams
If you look closely at most product organizations, the real bottleneck isn’t development.
It’s decision making.
Backlogs grow not because engineers are slow, but because teams struggle to continuously connect:
-
user feedback
-
product capabilities
-
strategic direction
-
delivery work
Information lives in different tools.
Insights get lost.
Discovery and delivery drift apart.
And thinking becomes fragmented.
From Backlogs to Thinking Systems
This is the problem space I’ve been exploring while building Outcomet.
Instead of managing static backlogs, the idea is to create a product operating system where thinking becomes continuous.
Where:
-
feedback reshapes capabilities automatically
-
strategy connects directly to discovery work
-
AI agents assist with heavy thinking tasks
Not replacing product managers.
But removing the cognitive overhead around them.
Less manual analysis.
Less backlog gardening.
More time for actual product discovery.
The Next Optimization Frontier
If the last decades optimized building software, the next decade might optimize something else entirely:
How efficiently we think about software.
AI doesn’t just accelerate coding.
It accelerates sense-making.
And the teams that learn how to structure, guide, and leverage that thinking will move dramatically faster than those still managing static backlogs.
A Simple Experiment
If you’re working with AI already, try this small experiment.
Track how often you ask yourself:
-
“Is this prompt worth it?”
-
“Should I run this again?”
-
“Do I really need this analysis?”
You’ll notice something.
You’re not optimizing compute.
You’re optimizing thinking.
Curious how AI is changing product work for you?
I’d love to hear what changed in the way you build, analyze, or make decisions.
And if you're interested in the system I'm exploring around this idea, you can take a look here:
