# How to Run a Pre-QC Pass That Actually Saves Senior Engineer Time **By Dan Cumberland** · Published May 19, 2026 · Categories: AI Strategy > Architecture time at the senior engineer level is the most expensive hour in your firm— and it gets spent finding missing dimension callouts that a junior... ## The Senior Engineer's Most Expensive Hour Architecture time at the senior engineer level is the most expensive hour in your firm— and it gets spent finding missing dimension callouts that a junior should have caught three weeks ago\. A pre\-QC pass, done right, returns that hour to the work it should be doing\. Most $20M–$100M AEC firms already run a version of this informally\. The principal flips through a 50\-sheet set the night before submission, marks up the obvious misses, and hands it back to the design team\. It works, sometimes\. It doesn't scale\. This article gives you a runnable framework— five steps, defined roles, clear handoffs, and a specific place where AI sits in the workflow\. Think of it as the operational layer that decides whether [AI implementation that fits an existing operational workflow](/services/ai-implementation) actually saves senior time or just adds another dashboard to ignore\. A pre\-QC pass is a structured designer\-led review of drawings against a checklist before they're handed to a senior reviewer for formal quality control\. Every senior reviewer hour you reclaim from clerical work has higher leverage than any junior hour— because juniors aren't authorized to sign off, and seniors are scarce\. Before walking through the framework, it's worth getting precise about what a pre\-QC pass actually is\. ## What a Pre\-QC Pass Is— and What Most Firms Get Wrong A pre\-QC pass is the structured designer\-led review of drawings against a checklist before they're handed to a senior reviewer for formal QC\. Different firms call this different things: designer self\-review, internal QA, peer check\. The function is identical\. Catch the clerical, coordination, and code\-reference errors before they consume senior reviewer attention\. Formal QC sits one step downstream\. According to the City of Los Angeles Bureau of Engineering's Project Delivery Manual[1](/blog/blog-architecture-time#ref-1), formal quality control is performed by senior personnel not directly involved with the project, at 30%, 60%, and 90–100% design milestones\. That's the convention public\-sector work codifies and most private firms follow informally\. ```html-table
Pre-QC PassFormal QC Review
Who runs itDesign team / designated peerSenior engineer or architect not on the project
WhenBefore each formal QC milestoneAt 30%, 60%, 90–100% design milestones
What it catchesClerical, coordination, code-reference errorsDesign intent, constructability, structural judgment
AuthorityIdentifies issues; cannot sign offStamps the deliverable
``` The mistake most firms make is collapsing the two\. Wallace Design Collective[2](/blog/blog-architecture-time#ref-2) describes the designer self\-review as scanning for spelling errors, missing graphics, broken section references, and incomplete details— and notes that organizing and summarizing the design philosophy and calculations dramatically speeds the senior reviewer's work\. That's the pre\-QC pass doing its job\. When firms skip the structure, the senior reviewer ends up doing both jobs at once\. Pre\-QC catches the clerical; formal QC catches the judgment\. Confusing the two is why most informal designer self\-review fails to save senior time\. The reason this distinction matters is economic\. And the economics are sharper than most firms recognize\. ## Why Senior Reviewer Time Is the Real Bottleneck Senior engineer review time is the bottleneck for one structural reason: they're the only role authorized to sign off, and they're the least available hours in the firm\. According to Monograph's analysis of architecture business benchmarks[3](/blog/blog-architecture-time#ref-3), Partners, Directors, and Sole Principals average roughly 46% billable utilization— compared to about 64% for Associates\. Every hour you reclaim from clerical review is high\-leverage time returned to design, business development, or client\-facing work\. That's the input side\. The output side is rework\. PlanRadar's 2025 consolidation of academic research[4](/blog/blog-architecture-time#ref-4) places construction rework cost between 4% and 10% of total project cost, with most peer\-reviewed studies clustering in that range\. Design\-related errors specifically contribute 1–9% of total project cost historically, declining to 1–2% in recent years[5](/blog/blog-architecture-time#ref-5) as BIM and digitization have matured\. The good news: QA/QC discipline measurably reduces this\. PlanRadar's QA/QC Impact Report[6](/blog/blog-architecture-time#ref-6) found that 56% of firms with consistent QA/QC processes keep rework below 5% of project budget, compared to only 37% of firms without such standards\. Structured review isn't a marginal improvement\. It's a measurable difference in how much rework reaches the field\. There's a parallel data point from outside AEC worth flagging carefully\. In a 2023 survey of 250 engineering leaders, primarily in manufacturing[7](/blog/blog-architecture-time#ref-7), 23% of engineering time was spent on non\-value\-added work— meetings, hunting for information, waiting on feedback\. The number isn't AEC\-native, but the pattern is recognizable\. Clerical drift consumes more senior engineering hours than most firms track\. In practical terms, the economic case looks like this: - Senior reviewer hours are scarce by definition \(46% billable for principals\)\. - Rework costs 4–10% of project budget; structured QA/QC keeps it under 5% in most cases\. - Reclaiming senior time from clerical review compounds: less rework downstream, more billable capacity upstream\. With the economics established, the next question is mechanical\. What does the pre\-QC pass actually look like end\-to\-end? ## The Pre\-QC Pass Framework— Five Steps A pre\-QC pass runs in five steps: drawing freeze, automated first\-pass review, triage by a designated reviewer, designer correction loop, and curated senior handoff\. Each step has a defined role, an artifact, and a decision point\. The whole pass runs in the days before each formal QC milestone\. It's the operational pattern behind [our guide to AI workflow automation](/blog/ai-automation-guide)— deterministic work moves to AI, judgment stays with people, and the handoffs are structured\. ```html-table
StepRoleArtifactDecision Point
1. Drawing freezeProject designerDated drawing set + summary of design philosophy and calculationsSet declared ready for review
2. Automated first-passAI drawing review toolAI flag report (raw)All deterministic checks run
3. TriageDesignated triage reviewer (mid-level engineer)Curated issue listPriority and validity of flags assigned
4. Designer correction loopProject designerRevision logIssues resolved or escalated
5. Senior handoffSenior reviewerJudgment review logSign-off or markup
``` **Step 1: Drawing freeze\.** The designer issues a snapshot of the deliverable for review\. This isn't dramatic— it's a dated set with a short summary of design philosophy and calculations attached\. Wallace Design Collective[2](/blog/blog-architecture-time#ref-2) flags this organizing step as one of the highest\-leverage things a designer can do\. When the senior reviewer eventually picks up the set, the front\-loaded summary lets them orient in minutes instead of hours\. **Step 2: Automated first\-pass review\.** The AI tool checks drawings for deterministic issues— dimensions, callouts, cross\-sheet coordination, code references, sheet template compliance\. CoLab Software[8](/blog/blog-architecture-time#ref-8) describes the pattern: specialized agents trained for specific failure points, not generic LLMs, catch basic errors so engineers can skip straight to the second\-pass judgment work\. Engineering\.com's independent coverage of the same tool[9](/blog/blog-architecture-time#ref-9) confirms the framing— an agent performs the initial review and flags issues; engineers focus on nuanced decisions\. InspectMind[10](/blog/blog-architecture-time#ref-10) describes a similar architecture: drawings and specifications analyzed together to surface coordination issues, missing information, and constructability problems before they become rework\. The pre\-QC pass runs immediately before each milestone formal QC, on the deliverables being submitted to senior reviewer\. **Step 3: Triage\.** A designated triage reviewer— typically a mid\-level engineer— takes the AI flag report, groups duplicates, dismisses false positives, prioritizes the rest\. This is the structural answer to false\-positive flood\. Without triage, every AI flag goes to the senior reviewer as raw output, and the senior reviewer ends up reviewing the AI's review\. Net\-zero, sometimes net\-negative\. A designated triage reviewer takes the AI report, prioritizes the flags, and routes only a curated list to the senior reviewer\. This is the structural answer to false\-positive flood\. **Step 4: Designer correction loop\.** The designer addresses the curated items\. Easy fixes get fixed\. Anything requiring judgment gets escalated with a note\. The revision log captures what changed and why\. **Step 5: Senior handoff\.** The senior reviewer receives a clean, summary\-documented deliverable plus the curated unresolved\-judgment items only\. They never see raw AI output\. They never spend time on clerical catches\. Their attention goes to the items that genuinely require their authorization— design intent, constructability tradeoffs, structural philosophy\. Vendor reports give a rough sense of speed compression\. Stru AI reports[11](/blog/blog-architecture-time#ref-11) that a 50\-sheet drawing set can be reviewed in 15–30 minutes versus 8–12 hours manually— treat that as a directional vendor claim, not an audited benchmark\. The savings are real; the magnitudes are vendor\-curated\. The AI step deserves its own treatment, because it's the most over\-promised and under\-specified piece of the workflow\. ## Where AI Belongs in the Pre\-QC Pass— and Where It Doesn't AI handles deterministic checks reliably in 2026— dimensions, callouts, cross\-sheet coordination, code references, layer and file conventions\. AI does not handle judgment— design intent, constructability tradeoffs, client\-specific preferences, structural philosophy\. The pre\-QC pass works when this line is drawn explicitly and enforced in the triage step\. ```html-table
AI handlesHumans own
Dimension and callout checksDesign intent
Cross-sheet coordinationConstructability tradeoffs
Code-reference verificationClient-specific preferences
Layer and sheet template complianceStructural philosophy
Spec-to-drawing consistencyStamp authority
``` Several tools sit in this category\. Mention them as enablers, not endorsements— each has a different focus and the AEC tooling landscape is still settling: - CoLab AutoReview[8](/blog/blog-architecture-time#ref-8) — multi\-agent first\-pass review with specialized agents per failure point - Stru AI[11](/blog/blog-architecture-time#ref-11) — multi\-agent drawing review aimed at speed compression - InspectMind[10](/blog/blog-architecture-time#ref-10) — plan check and drawing review with spec cross\-referencing - UpCodes Copilot[12](/blog/blog-architecture-time#ref-12) — code compliance specifically; scored 93% on a code accuracy benchmark in vendor testing The honest read on vendor metrics: direction is consistent across sources, magnitudes are vendor\-curated\. AI design review tools perform deterministic checks\. They cannot substitute for senior reviewer judgment on design intent or constructability tradeoffs\. Vendor claims of 93–97% error detection should be treated as directional, not as industry benchmarks— they're reported by vendors on curated drawing sets, not audited across the messy reality of sub\-consultant deliverables\. One precondition matters more than tool choice: CAD/BIM hygiene\. Real\-world drawings aren't vendor demo sets\. Inconsistent layer names, non\-standard sheet templates, and sub\-consultant files that don't match firm conventions all degrade AI flag quality\. Without input hygiene, the AI surfaces noise\. With it, the deterministic layer is fast and accurate\. McKinsey research[13](/blog/blog-architecture-time#ref-13) frames the macro case— AI can boost construction productivity by up to 20% through better planning, resource management, and design coordination\. That's the ceiling; the pre\-QC pass is one of the practical floors\. The framework is portable, but the activation cost is real\. The way to manage that cost is to start small\. ## How to Pilot a Pre\-QC Pass Next Monday Pilot the pre\-QC pass on one project, one discipline, one tool, one designated triage reviewer— for one milestone\. That's the minimum viable scope, and it's enough to generate the data you need to decide whether to scale\. **Monday\-Morning Pilot Checklist:** - **One project**, currently at or approaching the 60% design milestone— past conceptual chaos, before the construction\-document crunch\. The 60% gate is a clean signal per LA Bureau of Engineering convention[1](/blog/blog-architecture-time#ref-1)\. - **One discipline**, ideally the one with the most cross\-sheet coordination work in this project— typically structural or MEP\. - **One AI tool**, chosen for the discipline\. Don't shop the whole market; pick a single tool that fits the discipline and run it\. - **One designated triage reviewer**— a mid\-level engineer who has the technical fluency to dismiss false positives but isn't the project lead\. - **One milestone, two\-week timeline\.** The pre\-QC pass happens in the days before the 60% formal QC submission\. Senior reviewer approves the framework, then receives the curated handoff\. The senior reviewer's role in the pilot is structural, not operational\. They sign off on the framework, define what "curated" means for them, and receive the output\. They do not run the pre\-QC pass\. That's the whole point\. Before the pilot meeting, three things will trip up a pre\-QC pass faster than a bad tool choice— and they're worth knowing in advance\. ## Watchouts— False Positives, Knowledge Transfer, and CAD Hygiene Three problems will break a pre\-QC pass faster than tool choice: AI false\-positive fatigue, junior engineer knowledge\-transfer loss, and inconsistent CAD/BIM hygiene from sub\-consultants\. Each has a structural answer\. These are also the [hidden costs of AI projects to plan for in advance](/blog/hidden-costs-ai-projects)— predictable failure modes that show up after the contract is signed\. 1. **False\-positive flood\.** If AI surfaces too many false positives, the senior reviewer ends up reviewing the AI's review— net\-zero or net\-negative time saved\. **Mitigation:** the triage reviewer sits between AI output and senior reviewer\. Senior never sees raw AI flags\. Triage's job is to make the senior's input list shorter and higher\-signal than it would have been without AI in the workflow\. 1. **Knowledge\-transfer loss\.** When AI catches everything in first pass, juniors stop seeing senior markups— which is how they learned the trade in the first place\. **Mitigation:** preserve senior review of judgment items with the junior present\. Junior shadows the substantive review, not the clerical catch\. The trade is still teachable; the curriculum just moves up a layer\. 1. **CAD/BIM hygiene\.** Inconsistent layer names, sheet template drift, and sub\-consultant deliverables that don't match firm conventions all degrade AI output\. **Mitigation:** standardize file naming, layer conventions, and sheet templates before running AI in production\. Wallace Design Collective's organizing discipline[2](/blog/blog-architecture-time#ref-2) applies here too: clean inputs make every downstream step faster\. There's also an open governance question that doesn't have a public answer yet: E&O \(errors and omissions insurance\) and licensing implications of AI\-assisted review on stamped deliverables\. Senior reviewers still stamp\. AI doesn't\. But firm legal and E&O policy on AI tool use in the stamped\-deliverable workflow is worth a separate conversation— don't assume\. Once the pilot is running, the question becomes how to know whether it's working\. ## What to Measure Three metrics tell you whether a pre\-QC pass is working: senior reviewer hours spent per milestone, RFI volume during the corresponding construction phase, and rework cost on the closed project\. Track these on the pilot, compare to your last comparable project, and decide\. This is where [measuring AI success across operational pilots](/blog/measuring-ai-success) keeps the framework honest\. - **Senior reviewer hours per milestone** — the input metric\. If this number doesn't drop on the pilot relative to a comparable past project, the pre\-QC pass isn't working yet\. Iterate on the triage layer before scaling\. - **RFI volume during construction** — the downstream signal\. Fewer field questions on the post\-pilot project suggests the curated handoff is catching more before submission\. - **Rework cost as % of project cost** — the leading indicator\. PlanRadar's framework[6](/blog/blog-architecture-time#ref-6) aims firms at sub\-5% rework as the structured\-QA/QC bar\. That's the target\. If senior reviewer hours per milestone don't drop on the pilot, the pre\-QC pass is not yet working— keep iterating on the triage layer before scaling\. Set a baseline before piloting\. Compare against the most recent comparable project, not against industry averages\. Your firm's baseline is the only baseline that matters\. The methodology is portable\. The hardest part is starting— and the firms that get there fastest are the ones that pilot honestly rather than chase a perfect rollout\. ## Frequently Asked Questions ### What is a pre\-QC pass in architecture and engineering? A pre\-QC pass is a structured review by the design team \(or a designated peer\) against a checklist, before formal senior\-engineer QC review\. Different firms call it designer self\-review, internal QA, or pre\-QC— the function is identical: catch clerical, coordination, and code\-reference errors so senior review focuses on judgment\. The structure is what separates it from informal designer self\-review; without defined roles and artifacts, the senior reviewer ends up absorbing the clerical work anyway\. ### How much does rework cost an AEC project? Most academic studies cluster construction rework cost between 4% and 10% of total project cost[4](/blog/blog-architecture-time#ref-4)\. Design\-related errors specifically contribute 1–9% historically, declining to 1–2% in recent years[5](/blog/blog-architecture-time#ref-5) as BIM and digitization have matured\. Structured QA/QC discipline measurably reduces this— 56% of firms with consistent QA/QC keep rework below 5%, versus 37% without standards[6](/blog/blog-architecture-time#ref-6)\. ### Can AI replace senior engineer review? No\. AI handles deterministic checks— dimensions, callouts, code compliance, cross\-sheet coordination— reliably in 2026\. Judgment\-level review of design intent, constructability tradeoffs, and client\-specific preferences remains human\. The pre\-QC pass works because it draws this line explicitly: AI handles deterministic, humans own judgment, and a triage reviewer keeps the line enforced\. ### How long does AI drawing review take versus manual review? Vendor reports indicate substantial compression\. Stru AI states[11](/blog/blog-architecture-time#ref-11) a 50\-sheet drawing set can be reviewed in 15–30 minutes versus 8–12 hours manually\. Treat these as directional vendor claims, not audited industry benchmarks— and expect a triage layer to consume some of the savings, which is the price of keeping the senior reviewer's queue clean\. ### When in a project should the pre\-QC pass run? The pre\-QC pass runs immediately before each formal QC milestone— typically at 30%, 60%, and 90–100% design completion\. Public\-sector references such as the City of Los Angeles Bureau of Engineering's Project Delivery Manual[1](/blog/blog-architecture-time#ref-1) codify these as standard QC stage\-gates\. Running pre\-QC at each milestone keeps the senior reviewer's role focused on judgment at every gate, not just the final one\. ## Closing Architecture time at the senior engineer level is your firm's scarcest resource\. A pre\-QC pass— five steps, one triage reviewer, one AI tool handling the deterministic layer— returns that time to the work it should be doing\. The methodology is portable; the activation is the work\. The hardest part of running a pre\-QC pass isn't the AI tool— it's drawing the line between deterministic and judgment, and enforcing it in the triage step\. Pilot small\. Measure honestly\. Scale what works\. Dan Cumberland Labs helps AEC firms map AI to specific workflows like QA/QC— where the leverage is high and the failure modes are predictable\. If a 60% milestone is coming up on a project where this would land, [a conversation about implementation](https://dancumberlandlabs.com) might be worth having\. ## References 1. Bureau of Engineering, City of Los Angeles, "9\.3 Design Reviews — Project Delivery Manual, Chapter 9: QA/QC During Design" \(2024\) — [https://projectdeliverymanual\.engineering\.lacity\.gov/chapter\-9\-qaqc\-during\-design/93\-design\-reviews](https://projectdeliverymanual.engineering.lacity.gov/chapter-9-qaqc-during-design/93-design-reviews) 2. Wallace Design Collective, "Quality Assurance and Quality Control in Structural Engineering" \(2024\) — [https://wallace\.design/insights/quality\-assurance\-and\-quality\-control\-in\-structural\-engineering/](https://wallace.design/insights/quality-assurance-and-quality-control-in-structural-engineering/) 3. Monograph, "Architecture Business Benchmarks: Understanding and Improving Utilization Rates" \(2024\) — [https://monograph\.com/blog/unlocking\-utilization\-rates\-benchmarks\-for\-architects\-and\-architecture\-firms](https://monograph.com/blog/unlocking-utilization-rates-benchmarks-for-architects-and-architecture-firms) 4. PlanRadar, "Cost of Rework in Construction: Causes, Data & Prevention" \(2025\) — [https://www\.planradar\.com/us/cost\-of\-rework\-construction/](https://www.planradar.com/us/cost-of-rework-construction/) 5. PlanRadar \(consolidating Cnudde 1991, Burati 1992, Hwang 2009, Dougherty & Hughes 2014\), "Cost of Rework in Construction: Causes, Data & Prevention" \(2025\) — [https://www\.planradar\.com/us/cost\-of\-rework\-construction/](https://www.planradar.com/us/cost-of-rework-construction/) 6. PlanRadar 2025 QA/QC Impact Report, "Cost of Rework in Construction: Causes, Data & Prevention" \(2025\) — [https://www\.planradar\.com/us/cost\-of\-rework\-construction/](https://www.planradar.com/us/cost-of-rework-construction/) 7. CoLab Software / Global Surveyz Research, "23% of Engineering Time Spent on Non\-Value\-Added Work" \(2024\) — [https://www\.colabsoftware\.com/research/23\-of\-engineering\-time\-spent\-on\-non\-value\-added\-work](https://www.colabsoftware.com/research/23-of-engineering-time-spent-on-non-value-added-work) 8. CoLab Software, "AutoReview: AI that catches design issues before they slip through the cracks" \(2026\) — [https://www\.colabsoftware\.com/product/autoreview](https://www.colabsoftware.com/product/autoreview) 9. Engineering\.com, "CoLab launches AutoReview AI tool for design review" \(2025\) — [https://www\.engineering\.com/colab\-launches\-autoreview\-ai\-tool\-for\-design\-review/](https://www.engineering.com/colab-launches-autoreview-ai-tool-for-design-review/) 10. InspectMind AI, "Plan Check & Drawing Review" \(2026\) — [https://www\.inspectmind\.ai/checkers/](https://www.inspectmind.ai/checkers/) 11. Stru AI, "A Team of AI Agents Reviews Your Construction Drawings — Here's What They Catch" \(2025\) — [https://stru\.ai/blog/ai\-construction\-drawing\-review\-agent](https://stru.ai/blog/ai-construction-drawing-review-agent) 12. UpCodes, "UpCodes Copilot \| AI\-Powered Compliance Research Assistant" \(2026\) — [https://up\.codes/features/ai](https://up.codes/features/ai) 13. McKinsey & Company, "How the construction industry can boost productivity through technology" \(2024\) — [https://www\.mckinsey\.com/uk/our\-insights/the\-mckinsey\-uk\-blog/how\-the\-construction\-industry\-can\-boost\-productivity\-through\-technology](https://www.mckinsey.com/uk/our-insights/the-mckinsey-uk-blog/how-the-construction-industry-can-boost-productivity-through-technology) --- Source: https://dancumberlandlabs.com/blog/architecture-time/