I write about AI, product leadership, and strategy - exploring how technology can act as the glue that keeps vision and execution aligned.
Many of these posts first appeared on my LinkedIn page. Here, I’m curating them as a longer-term archive.
Articles:
Featured Posts:
Originally published on Linkedin
We did everything right—on paper.
We built an AI/ML-powered system to improve customer delivery promise outcomes—using predictive models to catch defects beforea or right when they happen.
We:
- Wrote an in-depth BRD
- Reviewed it with stakeholders
- Prioritized six high-potential use cases
- Built, tested, and backtested a Random Cut Forest model
- Launched and shipped
And two of those use cases truly landed.They identified critical defects, prevented prolonged CX issues, and delivered measurable business value. They still run today—now owned by another team. That work counted.
But the rest ?
- Some technically accurate predictions that didn’t lead to action
- And some “real” defects that didn’t matter in the broader business context
- Even when the model was right, the signal wasn’t useful—or usable
If I could go back, I wouldn’t have shipped all six.
And I wasn’t new to product—and I was working with a senior SDM and a Principal Scientist. We had horsepower, buy-in, and alignment. But I was over-eager. I pushed to land the full roadmap.
Looking back, I traded scope for clarity. And the team paid the price—in time, in complexity, and in missed opportunities.
What I’ve learned is this:✅ A model isn’t a product✅ Accuracy ≠ action✅ Detection without resolution is just noise
Today, I lead differently. I ask:
- I push for outcome clarity before technical execution
- I validate use cases through the lens of business impact
- I treat ML as an enabler—not the answer
Because building AI or ML-powered products isn’t about what can be predicted—it’s about what’s actionable, valuable, and worth solving.
This work was about four years ago—and one thing it did prove is that AI/ML can identify far more anomalies than humans ever could.If we rebuilt it today with modern tools, I believe we could drive better automation, SOP-level resolution, and root cause detection.
But the lesson still holds:
Business outcome is what matters.Not just the AI model.And not all anomalies are worth solving—if they don’t change what matters.
Originally published on LinkedIn
Not every product manager sets it. Some just ship the roadmap. But at senior levels, especially in places like Amazon, you’re not handed a blueprint. You create one.
That reminder came into focus recently during a mentorship conversation. My mentee jumped straight into solving a complex CX issue, without setting the vision or defining what “good” looked like.
I’ve been there. When the path is murky, it’s tempting to dive into action. So I shared a story from my own recent experience, a time I was asked to “look into” a long-standing problem with lead time.No charter. No team. Just a vague line in a doc.
This wasn’t the first time I’d stepped into a high-stakes, ambiguous space without a formal mandate.
Over time, I’ve built a simple approach to shaping vision from zero:
1. Understand the broader business context- what actually matters
2. Be honest about where things stand today
3. Separate signal from noise
4. Define the foundational building blocks
5. Lay out a direction that earns trust and alignment
In this case, I framed the opportunity, wrote the first strategy draft, and aligned stakeholders across different orgs, without formal authority. That work is now in motion, and I’m excited to see where it leads.
This is the kind of work I gravitate toward: foundational, fuzzy, high-impact.Because at a certain level, it’s not about delivering what’s already on the roadmap.It’s about shaping what deserves to be on it.
If you’re building from zero, always open to trading ideas, these are the challenges I love working through.