I argued against the AI. Then took the product to market.
A native plant nursery was managing hundreds of thousands of plants through spreadsheets, labels and one person's memory. I led Spriggo from product strategy through design, development and go-to-market.
- Role
- Founder-operator: product strategy, design, development and go-to-market
- Context
- Built at Do AI as a product venture, with a New Zealand native plant nursery as the pilot customer
- Project Type
- 0 to 1 product, B2B SaaS
- Timeframe
- October 2025 to present
- Status
- Live in production. Public launch August 2026.
- Product Strategy
- Product Design
- Development
- Go to Market
The operation depended on one person
The nursery grows hundreds of thousands of plants across species, batches, locations and lifecycle stages. Orders can be due next week or next year, while every batch changes independently as plants are propagated, moved, graded and sold.
Much of that operational knowledge lived with John, the nursery owner. The spreadsheets recorded stock, but they could not reliably show what was happening, what needed attention or what had already been promised to a customer.
Counting plants was not the first problem to solve
The initial opportunity was computer vision: point a phone at a bench and count the plants. It was an appealing AI feature, but it addressed one uncertain number inside a much larger information problem.
I argued that the product first needed a reliable operational layer. We scoped the initial release to inventory and sales, built around batches that follow plants from seed collection through to dispatch. The admin view gave the person running the business an overview; the worker experience brought the same system onto the nursery floor.
The market tested what we had treated as universal
The pilot nursery went live with real inventory in January 2026. The harder question was whether we had built a product or an unusually polished custom system.
Conversations at the national nursery trade day confirmed that the core model transferred: other nurseries also organised work around batches, inventory and sales. They also exposed where we had encoded one nursery’s habits too rigidly. Lifecycle stages, batch names and pricing structures varied in ways the first model did not allow.
One nursery’s detailed feedback moved through five pull requests in three days. The larger lifecycle issue required a proper architectural change rather than another exception for one customer.
Early evidence, with more to prove
These are early signals, not proof of product-market fit. The work shows how I make product decisions under uncertainty, test what transfers and respond when the market contradicts the model.