Written by
Zach Kelling
At
Thu Jun 08 2017
Techstars '17: What I Actually Learned
Hanzo went through Techstars in 2017. Nine years into building, here's what the accelerator actually taught me — not the PR version.
Techstars '17: What I Actually Learned
We got into Techstars in the spring of 2017. Nine years into building Hanzo — paying enterprise customers, a growing AI stack, a clear thesis about where commerce was going. I wasn't a seed-stage founder figuring out product-market fit.
Why apply? The network. Specifically: the enterprise relationships that a Techstars voucher compresses into months instead of years.
If you're selling AI infrastructure to enterprise buyers, you need someone who can say "these people are real" in a room you're not in. That's what the Techstars brand does. Thirteen weeks in Denver to get that.
The Most Useful Feedback
The question mentors pushed on hardest: If a large customer leaves, what happens to your model quality?
This forced me to articulate something I'd been assuming: the Hanzo models were trained on aggregate behavioral data across hundreds of clients. A single large client was maybe 5-15% of the training corpus. Losing them would matter, but the models wouldn't degrade catastrophically. And critically — diversity of training data matters more than volume above a certain threshold. The small clients in diverse categories were, paradoxically, more valuable to model quality than a single large one in a single category.
I've been making this argument to enterprise buyers ever since. Techstars is where I learned to make it clearly.
Demo Day
We presented the AI-native commerce thesis in September: platforms that learn from every transaction compound against the ones that don't, and the compounding curve is already steep.
The follow-on round closed within three months. That funded the multi-region infrastructure buildout, the ML v2 work, and the enterprise expansion.
What I'd Tell Myself
Apply earlier. The relationships formed in a cohort get more valuable over time — the people you meet become more useful as they build more. The Techstars '17 cohort is more useful to me in 2026 than it was in 2017.
The accelerator didn't make Hanzo. Nine years before it did that. What it did was compress relationship timelines. That's the actual value.
Hanzo is a Techstars '17 company. Zach Kelling is the founder.