After VideoHQ.ai validated demand for opinionated AI workflows, JamaKit.ai extended the bet: a toolkit small teams could use to compose their own AI workflows without engineering support.
Teams kept asking for VideoHQ-style automation in adjacent domains. Building a separate product for each was untenable; the platform play was to ship the primitives.
- Define a composable workflow primitive
- Ship a launch surface non-engineers could use
- Reuse the prompt library across products
- Create a clear bridge from VideoHQ users into JamaKit
Interviewed VideoHQ power users and prospective JamaKit users to map the workflows they were already hacking together with spreadsheets and standalone AI tools.
Modeled the product around three primitives: input source, AI step, output destination. Everything else was opinionated defaults that users could override.
- Wrote the product specification and primitive contracts
- Designed the workflow builder with engineering and design
- Curated the launch templates from real user workflows
- Coordinated cross-product positioning with VideoHQ.ai
Resisting feature creep, every interview surfaced another step type, and the primitive set had to stay small enough to remain composable.
JamaKit.ai launched as a modular AI toolkit, reused the VideoHQ prompt library, and gave small teams a way to compose their own AI workflows without engineering.
- Primitives beat features when you're building a platform
- Templates are the on-ramp; primitives are the product
- Cross-product positioning is a PM responsibility, not a marketing one