Building with Intention
Over the past year, the way we build has fundamentally changed.
As a PM, I’ve watched timelines compress in ways that didn’t feel possible before. Things that used to take weeks such as prototypes, workflows, even early-stage products; can now be stood up in a day with AI.
On paper, that’s a massive win.
In practice, it’s introduced a new problem:
We’re building faster than we’re thinking.
In our last blog, we discussed the paradigm shift from “Can you?” to “When will you?” and this follows closely behind. With speed, comes shifts in ways that we work - how will you change your perspective?
Speed Exposes Weak Thinking
One of the biggest shifts I’ve noticed is how easy it is to mistake motion for progress.
AI gives teams the ability to generate very quickly; sometimes to their detriment. Features, flows, even entire experiences may feel not well thought through. When something works on the surface, it creates a false sense of completion.
I’ve seen teams get to something that looks like a finished product without ever answering the most important questions:
Does this solve a real problem?
Does this actually work for a user outside of our own context?
Would anyone come back and use this again?
Before AI, time acted as a forcing function. You had to think through these things because building was expensive.
Now, that friction is gone.
And without replacing it, you end up accelerating toward the wrong outcome.
I Start With the End State
The way I’ve adjusted is simple in concept, but requires discipline in practice.
I don’t start with what we can build.
I start with what success looks like when we’re done.
Not in vague terms like “launch the product” or “ship the feature,” but in very specific, observable outcomes.
Things like:
A user can complete the core flow without guidance
The output is something we trust—not just something that runs
There’s clear evidence of value (time saved, better decisions, repeat usage)
Once that end state is clear, everything else becomes a process of working backward.
Milestones Became My Operating System
In a world where AI can generate almost anything, milestones have become my primary guardrail.
But not all milestones are useful.
I’ve learned to avoid milestones that measure activity:
“Frontend complete”
“API connected”
“Feature built”
Those don’t tell me if we’re actually making progress.
Instead, I focus on milestones that validate something real:
The first time a real user uses the product without help
The moment our data source is live and no longer mocked
When recommendations or outputs feel meaningfully accurate, not just anecdotal
When the product creates value without me needing to explain it
That’s when something shifts from being built to being real.
In the world of AI - that matters more than ever.
A Real Example: Building an AI “Budtender”
Recently, I’ve been building an AI-powered “Budtender”—a system designed to recommend cannabis products based on user preferences like desired effects, experience level, and format.
With AI, I was able to get a working version of this very quickly.
I had:
A frontend experience
A recommendation endpoint
A scoring system
Even structured product data residing in a stood-up Inventory System
From the outside, it looked like a functional product.
But when I stepped back, I realized something important:
I hadn’t actually proven anything yet.
The recommendations existed—but were they good?
The system returned results—but were they trustworthy?
The flow worked—but would a real user understand it?
That’s when I shifted from building features to defining milestones.
Instead of asking, “What’s next to build?” I asked:
Can a user go through the flow without explanation?
Are the recommendations actually personalized—or just formatted well?
Is the product data real and reliable (not mocked or incomplete)?
Would someone trust this enough to act on it?
Those questions changed how I built.
For example, integrating a real product source wasn’t just a backend task: it became a milestone:
“Live data is flowing, and recommendations reflect real inventory.”
Getting one recommendation “right” mattered more than generating ten that looked polished.
Because that’s what moved the product forward.
AI Increases Output, Not Understanding
One of the more subtle challenges with AI is that it always gives you something.
You rarely hit a wall. You rarely get stuck in the traditional sense.
But that also means you don’t get the natural signals that something is off.
You can build an entire system that works technically, but doesn’t work at all in practice.
So I’ve had to change the questions I ask.
Not:
“Did we build this?”
But:
“What did we just prove?”
Every milestone needs to answer that.
Velocity Still Matters, Direction Matters More
To be clear, I’m not advocating for slowing down.
If anything, the expectation for speed is higher now.
They’re using AI to:
Rapidly test assumptions
Explore multiple approaches in parallel
Validate ideas earlier in the lifecycle
Kill weak concepts before they scale
The advantage isn’t just speed: it’s faster learning cycles.
In product, learning compounds much more than output does.
How I Think About Building Today
When I’m kicking off something new, I anchor the team around a small set of milestones that would make the product feel real.
Not finished. Real.
Typically, that looks like:
A user can engage with it without context
The system produces an output we trust
There’s a clear signal of value
It works in an environment that isn’t controlled
Everything we build is in service of hitting those points.
AI is just the tool that helps us get there faster.
The Shift That Matters
The biggest change I’ve made as a product leader isn’t adopting AI.
It’s becoming more intentional about how we build with it.
Because AI has removed a lot of the friction that used to force clarity.
And without that friction, it’s easy to move quickly without ever aligning on what actually matters.
The teams that will win in this environment aren’t the ones who can build the fastest.
They’re the ones who:
Define success clearly
Measure progress through meaningful milestones
And use AI to accelerate learning, not just their output
Speed is no longer the differentiator.
Intentionality is.
Leave a comment below, excited to talk to you all about this one!
Jake Smith
