Concepts

The expert panel

Where the interview questions come from, why they stay consistent, and how Discovery Mode teaches the system a scope it has never seen.

DarkBird's interview questions are maintained by a panel of domain experts across Camera/Surveillance, Network Infrastructure, Access Control, AV, Structured Cabling, Wireless, and Project Management. The AI draws on this expert-built library every time it scopes a project — so you get consistent, technically precise questions without any setup required.

Expert panel model

  • Questions are curated and maintained by DarkBird's domain experts — not generated on the fly
  • Each domain expert knows their area deeply and collaborates with other experts when a project spans multiple trades
  • Questions are stable: the Bird asks the same high-value questions consistently across projects, so your team builds repeatable scoping muscle memory

Discovery Mode and automatic learning

When the AI encounters a scope it has not seen before, it enters Discovery Mode and asks clarifying questions to understand the new service. Those questions are captured automatically. When the same pattern appears across multiple projects, it is promoted into the library — no admin action needed.

  • Discovery questions are saved behind the scenes — your team just answers them naturally
  • Frequently asked discovery questions are automatically elevated into the permanent question library
  • Auto-promoted questions take effect in your interviews right away, and are flagged for DarkBird's review so we can keep the shared question library high-quality over time
Note

There is no manual Interview Setup step. The question library is managed by DarkBird on your behalf. If you have a domain-specific question that should always be asked, contact DarkBird support to have it added to the expert panel.

How knowledge and expert profiles work together

  • Every scoping interview injects the relevant Knowledge Category, the per-category Expert Profile, and the org-wide Expert Profile before the AI responds
  • Card suggestions incorporate your org's estimation patterns from the Knowledge Categories
  • Expert Profiles shape the AI's follow-up questions — it probes the way your senior engineer would, not generically
  • SOW generation references your house rules and preferred phrasing from the Expert Profiles
  • Calibration insights from closed project reviews are merged back into the Knowledge Categories automatically