The best superintendents we know all run some version of a morning loop. They wake up early, walk the slab, scan yesterday’s reports, get ahead of the day. The bad ones spend 90 minutes in the trailer chasing email. The difference is not work ethic. The difference is that the great ones have a structured loop and the bad ones have an inbox.
Field-grade AI compresses the great super’s loop into 12 minutes — and gives the average super the structure to become great. Here is the loop, exactly.
Minute 0–2: Voice-first sync
Phone in hand, walking the slab. The super dictates: "Yesterday we got framing done on the east side, MEP rough-in is two days behind on level 3, the steel delivery slipped to Thursday." That voice note becomes the start-of-day update on every relevant project record. No typing.
Minute 2–4: RFI triage
The AI shows the three RFIs the super needs to action today, ranked by schedule impact. For each: a draft response based on the project graph, the engineer to route it to, and a one-tap "send" or "edit then send". A great super clears morning RFIs before the field even starts.
Minute 4–6: Slip risk scan
AI surfaces the five activities most likely to slip in the next two weeks, with the reason — vendor lead time, dependency depth, weather. The super taps "acknowledge" or "mitigate"; the mitigation kicks off the right notification chain.
Minute 6–9: Crew alignment
The super reviews the AI-drafted today plan: who is on which scope, where conflicts exist (overlapping trades, gear bottleneck), and what the high-leverage moves of the day are. Edits in seconds; publishes to crew leads.
Minute 9–11: Toolbox talk
ToolboxTalksAI generates today’s topic from yesterday’s site conditions and the next 48 hours of work. The super reviews, optionally re-records in their own voice, posts.
Minute 11–12: Daily huddle prep
The huddle now runs in six minutes because everyone walks in already aligned to the same numbers. The super spends the huddle on judgment calls, not status updates.
Why this works
- It is voice-first, so the super stays on the slab.
- It produces real artifacts (RFI sends, mitigations, crew assignments) — not just summaries.
- It compounds: every loop entry is reusable context for tomorrow’s loop.
- It is offline-first; reception failure does not break the cadence.
30-day measurement
Measure two metrics for 30 days: average RFI cycle time and weekly slip-risk score. Both will improve in the first two weeks if the loop is run with discipline. If they do not, the AI is failing one of the four artifacts — diagnose which one and tighten it.
Run the loop for a quarter and the field stops being a place where information is lost. It becomes a place where information is generated.