What your organization can own
Each engagement is set up to leave your organization owning the layers in its agreed scope.
| Layer | You can own | Still depends on |
|---|---|---|
| Data pipelines | Intake and validation code, data contracts, run history | Your source systems and their export formats |
| Analysis | Versioned transformations, indicators, evaluation sets | The database platform you choose |
| Models | The interface your applications call, model choice, prompts, evaluation results | Open-source license terms, operating system, drivers |
| Records | Export and recovery procedures for every data store created | The storage you supply |
| Knowledge | Architecture records, operating procedures, trained staff and fellows | Staff retention |
An engagement can include a written reason and exit path for each build-or-buy choice.
The lead's experience at his employer
- Not another pilot. At his employer, Justin Jones leads the team that runs AI tools in production across a multi-company portfolio in media, retail, hospitality and financial services.
- Not paying by the row to move your own data. Pipelines his team built keep append-only raw history, and the company owns the ingestion code and data contracts.
- Not adding a SaaS seat for every workflow. His team replaced a SaaS time and billing tool with a company-owned finance platform whose change records name the person who acted.
- Not betting on one AI vendor. The AI coding tools his team built hand a task from one agent provider to another.