Case Study: SheetIntel Saved 12 Hours on a Caribou Coffee Plan Review

The Situation

A commercial general contractor bidding a Caribou Coffee tenant improvement in a suburban retail strip center had two days to review a 187-sheet plan set and submit a competitive bid. Their senior estimator had a realistic capacity for about 4 hours of plan review on this project — already competing with two other active bids.

A 187-sheet commercial plan set reviewed in 4 hours means roughly 1.3 minutes per sheet. An experienced estimator can move that fast on straightforward projects. A food service buildout is not a straightforward project.

Food service TIs are among the most coordination-intensive work in commercial construction: commercial kitchen exhaust routing through occupied ceiling plenums, Type I hood clearances relative to structural framing, grease interceptor sizing and location, health department requirements layered on top of standard building code, and high-voltage service for specialty equipment where owner-furnished specs may not match the electrical drawings. Every one of these is a potential bid day ambiguity or post-contract change order.

The estimator had done Caribou Coffee buildouts before. That familiarity was an asset — and a risk. Experienced estimators who know a client's prototypical plans sometimes fill in details from memory that may not actually be present in the drawings in front of them.

How They Used SheetIntel

The estimator uploaded the 187-sheet PDF to SheetIntel before starting their manual review. The goal was to get SheetIntel's output first — using it as a pre-read to direct where manual attention should go, rather than treating it as a replacement for review.

SheetIntel processed the plan set and returned a structured report in under 15 minutes: a trade-by-trade breakdown of flagged items organized by conflict type (coordination conflicts, specification cross-references, scope attribution gaps, and dimension/detail inconsistencies) with specific sheet and detail callouts for each flagged item.

The estimator then spent their 4 hours working through the flagged items with the context SheetIntel had surfaced — verifying conflicts, making judgment calls on which to include in the bid letter, and determining which needed formal RFIs before bid submission. That's a fundamentally different use of 4 hours than undirected review: human judgment applied where it matters, not spread thin across 187 sheets hoping to catch something.

Three Conflicts That Would Have Cost Money

SheetIntel flagged three cross-discipline conflicts that the time-constrained manual review would not have caught:

1. Kitchen exhaust duct routing vs. structural steel

The mechanical drawings routed the Type I exhaust duct from the commercial kitchen through a ceiling plenum bay where the structural drawings showed a steel beam at the same elevation. The conflict was only visible when overlaying the mechanical and structural plans — not something an estimator reviewing one discipline at a time would catch in a compressed timeframe.

In the field, this would have surfaced at rough-in as a duct reroute requiring new penetrations, potential fire-rated assembly work at the new duct path, and sheet metal coordination changes. The early RFI let the architect resolve the routing in the permit set before the job started.

2. Grease interceptor: interior vs. exterior location conflict

The plumbing drawings showed an interior grease interceptor in the back-of-house utility area. The civil drawings indicated the exterior interceptor at a different connection point with a different service run length. The two drawings described different scopes — the GC would have priced one without knowing they were bidding something different than what the civil engineer had drawn.

This type of conflict — plumbing and civil describing the same system differently — is common in food service tenant improvements where the MEP and civil engineers often work off different versions of the architectural floor plan. An RFI before bid day forced clarification on which drawing governed, protecting the GC's scope definition from the start.

3. Electrical panel schedule vs. owner-furnished equipment

The owner-furnished equipment schedule listed a 3-phase espresso machine requiring a 60A circuit. The electrical panel schedule included a 30A circuit for that equipment location. The two documents were inconsistent — and since the electrical subcontractor bids off the panel schedule, not the equipment schedule, the 30A circuit would have been installed and the field change would have come later.

A panel modification after rough-in — adding breaker capacity, potentially upgrading the service run — typically runs $3,200–$4,800 on a commercial TI. The GC included a scope allowance in their bid letter and won the job with a complete picture of their electrical exposure.

The Outcome

The GC submitted RFIs on all three items before bid day. Where RFI responses wouldn't arrive before bid submission (the panel conflict), they included a documented allowance in the bid letter, which protected margin without making the bid uncompetitive.

They won the job. Post-award, the architect's RFI response on the exhaust duct conflict required a reroute that added $2,100 in mechanical work — scope that was captured in the bid because the GC had flagged it proactively and included a clarification.

The estimator's post-bid assessment: finding and documenting these three conflicts through undirected manual review would have taken 12–16 hours of careful cross-discipline work. SheetIntel surfaced them in 15 minutes, directing 4 hours of human review to exactly the right places.

Why Food Service TIs Are High-Risk for Bid Errors

The three conflicts found in this plan set aren't unusual for food service construction. They reflect structural patterns in how these projects are designed and documented:

MEP density is high. A 2,800 SF coffee shop has more mechanical, electrical, and plumbing systems per square foot than most commercial occupancies. Kitchen exhaust, makeup air, refrigeration, grease waste, high-amperage service, and low-voltage technology systems all compete for the same ceiling plenum. Coordination conflicts are structurally likely, not the exception.

Owner-furnished equipment creates a documentation gap. Corporate restaurant clients often specify equipment on schedules that are prepared separately from the base building drawings. When the equipment schedule and the electrical drawings originate from different design teams on different timelines, mismatches are common and frequently go undetected until rough-in.

Prototypical plans create false familiarity. GCs who build for the same franchise client repeatedly develop pattern recognition about what these projects typically include — which is valuable, but creates blind spots when a specific location's plan deviates from the prototype. AI review doesn't rely on pattern recognition; it reads what's actually on the sheets.

Generalizing the Result

The time math on this job — 15 minutes of AI review directing 4 hours of human attention versus 12–16 hours of undirected manual coverage — scales across every bid in a GC's pipeline. The compounding value isn't one saved change order; it's thorough review on every job, not just the ones where bid timing happens to allow it.

Pre-bid plan review is the highest-leverage point in the construction lifecycle for catching problems cheaply. A scope conflict found at bid is a clarification. The same conflict found at rough-in is a change order. The same conflict found at closeout is a dispute. The earlier the catch, the lower the cost — typically 10–100× lower at bid vs. construction.

SheetIntel doesn't replace the estimator's judgment on which conflicts matter, how to price ambiguous scope, or which clarifications to put in the bid letter. It front-loads the systematic work of finding those conflicts so that judgment goes where it's needed.