Persistent Agents for the Built Environment Visual-story transcript Runtime: 1:17 Voice: Samantha, US English, 200 words per minute 1. A small revision Monday morning. A structural opening moves. A mechanical route may clash, a fabrication package may be outdated, and the programme may already be exposed. 2. The continuity problem Each team knows a part. The problem is continuity: what was known does not arrive together where the next decision is made. 3. The idea A persistent agent carries one bounded responsibility across time. It waits, notices a relevant event, restores accepted context, and returns with the evidence needed next. 4. When the revision arrives When a revision arrives, it traces what changed, finds affected relationships, identifies the accountable owner, and prepares a review. It records the decision, then waits again. 5. Bounded roles This pattern can follow work packages, approvals, infrastructure dependencies, or a field across a season. Each agent has one role and a clear authority boundary. 6. Authority People remain accountable. An agent may observe, analyse, recommend, or prepare. Engineers, operators, project leaders, and public authorities still own consequential decisions. 7. The quiet engineering problem Persistence has a cost. Memory must be selective, sourced, permissioned, and time-aware. Without budgets and expiry, an agent can become noisy, expensive, or consistently wrong. 8. A responsible beginning Start with one recurring obligation, visible evidence, a named owner, and a measurable outcome. Read the full perspective in the BIMSAARTHI Community. Production note The screen uses short reinforcement phrases. The selectable caption track divides this narration into brief movie-style cues.