Project vs hire vs hybrid
Two teams can both say “AI build” and still need completely different engagement shapes. Start with who owns the finish line.
| Model | Best when | Weak when | You get |
|---|---|---|---|
| Project delivery | Scoped outcome and written acceptance | Priorities change every week | A finish line and milestone demos |
| Dedicated hire | Continuous discovery inside your cadence | No clear owner on your side | Capacity in standups and backlog |
| Hybrid | Need a first slice, then iteration | You want only a one-off dump of code | Sprint to ship, then embedded team |
Project delivery fits a defined outcome
MVP agent, migration, or a scoped admin rebuild—clear acceptance criteria and a finish line favor fixed or milestone delivery.
- You can describe done in writing (and disagree in review, not mid-sprint).
- Dependencies and environments are known enough to estimate.
- Success is a shipped slice, not “keep exploring forever.”
Dedicated hire fits continuous discovery
When priorities move weekly and you need people inside your standups, dedicated engineers keep context and ship incrementally.
- Roadmap items compete weekly; context loss costs more than a fixed SOW.
- You want named engineers on your tools and ceremonies.
- AI evals and UX need repeated loops after the first release.
Hybrid is common for AI
A short discovery/build sprint, then embedded capacity to iterate on evals and UX, often beats all-or-nothing choices.
- Scope a first production slice with acceptance criteria.
- Ship with evals, logging, and a human checkpoint where risk is high.
- Keep dedicated capacity to tune prompts, retrieval, and UX from real use.
- Review the engagement shape monthly—expand, hold, or finish cleanly.
