For each productization candidate: should SmartOne build it internally, buy a tool and integrate it, partner with a platform player, or license it? The decision framework and the specific recommendations per candidate.
The 70/30 split refers to a principle in platform design: roughly 70% of the platform value should come from capabilities SmartOne already has (the open-source tool stack, the annotation methodology, the QA pipeline), and roughly 30% from new capabilities that need to be added. This memo operationalizes that split into specific build, buy, and partner decisions.
| Capability | Build (internal) | Buy (vendor tool) | Partner (integration) | License (white-label) | Recommendation |
|---|---|---|---|---|---|
| [e.g., Pre-labeling AI assist] | [Cost / time / IP ownership] | [Labelbox SDK, Scale API, Encord] | [Snorkel integration] | [N/A] | [Buy: integrate Labelbox SDK] |
| [e.g., QA dashboard UI] | [Existing internal tool + UI layer] | [Off-the-shelf BI tool] | [N/A] | [N/A] | [Build: wrap existing tool] |
| [e.g., Annotation ontology engine] | [SmartOne already has this] | [N/A: already built] | [N/A] | [License to others: not initial scope] | [Build on existing asset] |
[The final section of this memo will be a three-layer stack diagram: existing SmartOne capabilities (bottom layer, already built), integrated vendor tools (middle layer, buy or partner), and new UI and customer-facing layer (top layer, the packaging investment). Each layer will have a cost estimate and a time-to-production estimate.]