Stashed: Built with AI. Made in India.

India's public transport is not gentle. Buses are packed. Metros are slippery. Anyone who's carried a backpack on the Chennai local knows the specific misery of a sweaty bag rubbing your back at 8am in March.

We wanted to solve the crossbody problem — a bag that sits flat against your body, doesn't catch on turnstiles, and doesn't scream “tourist.” The removable Velcro front panel was the differentiator: swap the look, keep the structure.

The unfair advantage: our family's B2B arm SN Bags has been manufacturing bags for schools and corporates for years. We had the factory relationships, quality control knowledge, and the credibility to get samples made at a scale most first-time founders can't access. We decided to point that infrastructure at a consumer product.

Three days of terrible names. UrbanKarry, BagWala, KarryCo. We went through a whiteboard's worth of ideas that all sounded like dropshipping stores from 2019.

We turned to Claude. The prompt was simple:

Suggest 20 brand names for a Made-in-India bag brand targeting metro commuters. Names should be short, modern, easy to spell.

Twenty names came back in seconds. Most were forgettable. One clicked immediately: Stashed. Short. Functional. The vibe of something urban. You stash your essentials. You stash it and go.

We checked availability, ran it past three people whose opinions we trust, and moved. Total naming process: 4 days. AI saved us from another 4 days of terrible options.

Before any physical sample existed, we needed to see the bag. Nano Banana became our visualization tool. The first 12 prompts were garbage — generic bags on white backgrounds that looked like AliExpress listings.

The refinement process took two days and a lot of iteration: adding material descriptors, lighting references, context cues. We were training ourselves to describe a product that didn't exist yet in enough detail that an image model could render it.

The prompt that finally worked placed the bag in an Indian metro context — tiles, soft window light, a commuter aesthetic. Three people DM'd asking where to buy before the product existed. That was the green light.

We used Claude to map manufacturers in Tamil Nadu and Gujarat. The AI gave us a starting list of names, regions, and known specializations — a research foundation that would have taken a week of manual Googling to assemble.

The SN Bags knowledge meant we could filter fast. We knew what MOQs were realistic (under 100 for samples), what fabric mills were credible in Tamil Nadu, and which questions to ask that separate serious manufacturers from middlemen.

Real problems AI couldn't solve: minimum order quantities that tripled on a WhatsApp call, sample mismatches where the color arrived 3 shades off, and one supplier who simply stopped responding after we paid for a sample.

What AI helped with: research, RFQ drafts, spec sheets, comparison matrices. What required phone calls and factory visits: everything else.

The gap between an AI render and a physical product is humbling. The first sample arrived and the zipper was in the wrong place — not wrong by the manufacturer's standards, but wrong for how a commuter actually reaches into a crossbody bag while standing in a moving metro.

The Velcro panel attachment was also wrong — it needed to be stronger for repeated daily use. The fabric felt stiffer than the cordura references we'd sent. AI can describe materials. It can't tell you how a zipper pull feels after 200 uses.

We redesigned after the sample: moved the zipper, changed the Velcro grade, and added a phone pocket accessible while wearing the bag. Three more sample rounds followed. None of this was on the AI prompt. All of it required hands on the product.

We built a waitlist before the product was finalized. AI-generated posts describing the bag's features — written from our product brief — started pulling real DMs. People wanted to know where to buy.

The first Instagram post was a render, not a photo. The response was the same: “This looks real, where can I buy?” — before it existed. That confirmed we had something worth manufacturing.

The pre-launch strategy was simple: show the product honestly, describe the problem it solves, and let the interest confirm the market. No paid ads. No influencer seeding. Just a clear product with a specific audience and AI-assisted content that didn't feel AI-generated.

First batch was small — intentionally. We didn't want to over-commit to a product we hadn't fully stress-tested with real customers.

First sales came through DMs and direct payment links. No ecommerce integration at first — just WhatsApp orders and UPI. Scrappy, but it shipped.

First return: a customer found the zipped inside pouch confusing — they didn't realize it was there and thought the bag was smaller than listed. That feedback became a product description update and a photo change.

Real numbers: not massive. Not viral. A brand doesn't get built on a first batch. What it gave us was proof, feedback, and a reason to keep going.

Stashed launches March 31, 2026. Less than 6 months from idea to first batch shipping. The product is better than the first sample. The manufacturing process is tighter. We understand the customer a little more with each order.

What AI still can't replace: the trust built through a factory visit, the manufacturing relationships that took years to develop at SN Bags, and the 4am anxiety that comes with running a physical product business. That part is still entirely human.

What's next: the Velcro ecosystem. Swappable front panels — different colors, different materials, limited edition collabs. The bag becomes a platform, not just a product. AI helps us prototype panel ideas in hours rather than weeks.

Every workflow you've read in this case study is the exact workflow we teach in 2BFT Academy. No theory. Just the actual process.