Construction Admin Is the Feedback Loop Your Design Team Never Got

What Construction Administration Actually Is (and What It Produces)

Construction administration is the architect's formal services during the construction phase— site inspections, RFI review, submittal approvals, change order documentation, punch list management, and certificate of substantial completion— as defined in AIA Documents B101 (Owner-Architect agreement)³ and A201 (General Conditions of the Contract for Construction)⁴.

B101 lists CA as one of the five phases of Basic Services, alongside Schematic Design, Design Development, Construction Documents, and Procurement³. A201 governs the Owner-Contractor-Architect relationship during construction itself and defines the RFI, submittal, and change order processes the architect reviews⁴. RFIs are triggered when the field encounters missing, incomplete, or incorrect information in the project documents⁵.

That's the procedural view. The design-feedback view is where the value lives. CA generates three categories of signal about design quality: questions the drawings should have answered (RFIs), gaps that cost money to fix (change orders), and defects discovered at handover (punch list items).

Knowing what CA is supposed to produce is one thing. Knowing what most firms actually do with that output is where the gap shows up.

Why Most Firms Under-Resource CA — and Discard the Data

Most architecture firms under-resource construction administration for structural economic reasons. Labor accounts for over 75% of an architecture firm's operating costs⁶. CA is typically billed hourly rather than as a fixed fee. And 75% of firms in the AIA's 2024 survey have fewer than 10 employees⁷— meaning a dedicated CA team is rarely affordable. The data CA generates is real; the systems to learn from it are missing.

Labor is 75% of an architecture firm's operating costs, which means under-staffing CA isn't a strategy failure— it's an economic survival reflex. Approximately 70% of change orders in construction projects originate from design errors and omissions, not from field conditions², which means most CA data is feedback about something the firm can actually fix.

The consequences show up in the response data. RFI first reply averages 6.4 days, with a median of 9.7 days⁸. Nearly 1 in 4 RFIs receive no reply at all⁸. A firm runs 200 RFIs on a $40M project, files every response, and never asks which detail types generated the most questions. What gets discarded isn't the individual response— it's the pattern across responses:

• Which detail types consistently generated questions
• Which change orders traced to the same coordination weak spot
• Which punch list items repeated across the firm's last five projects

This isn't a bad-actor problem. It's a missing-systems problem. The economic constraints are real. The thing worth asking is what those discarded data points would have told you if you'd looked.

What the Data Would Tell You — and Why POE Adoption Is Still 3%

Construction administration data reveals recurring design weaknesses— detail types that consistently generate RFIs, specification gaps that drive change orders, finish or coordination issues that dominate punch lists— yet only 3% of architecture practices in UK research regularly conduct formal post-occupancy evaluation to close the loop after the building is in use⁹.

RFI patterns are the firm's free design audit; most firms never run the audit. Only 3% of architectural practices regularly conduct post-occupancy evaluations on their work, leaving 97% to design the next building with no formal evidence about whether the last one worked.

What each signal type reveals when patterns are extracted:

• RFIs reveal which detail types and specification sections cause repeated confusion across projects
• Change orders reveal where scope estimation and consultant coordination consistently fail
• POE reveals how the building performs in use— energy efficiency, occupant comfort, functional success against design intent¹⁰

POE is the only one of the three that captures performance after the building is occupied. And it's the one almost nobody does. PlanMan's UK data shows 50% of firms never conduct POE for operational energy measurement¹¹. HMC Architects describes the pattern bluntly:

"Architecture design firms rarely engage closely with building performance which they have designed or built, and disconnect completely from the building once the project is finalized, missing either the chance of learning from previous mistakes or celebrating success and innovation."¹²

The peer-reviewed research is no kinder. A 2014 survey of industry practice found that current approaches to detecting and correcting errors significantly hinder learning from previous experiences— design-construction feedback loops are systematically broken in industry practice¹³.

Worth being honest: the 3% figure is from UK research and likely directionally correct for the US, not literally identical. The broader pattern— that the data exists and firms don't systematize it— is well-documented across both markets. The problem isn't that firms don't see the data. It's that systematizing it has historically required dedicated staff most firms can't afford. That math has changed.

AI Tools Are Quietly Solving the Capture Problem

Three categories of AI-enabled tools— visual jobsite documentation (OpenSpace, Buildots), construction project-data copilots (Procore Copilot), and lessons-learned database systems— have made the underlying CA-feedback workflow affordable for firms that could never staff a dedicated knowledge-management team. The technical bottleneck is now lower than the cultural one.

AI doesn't change the design judgment. It changes what's economically feasible to learn from each project. The point isn't to automate architecture. It's to give your CA team intellectual augmentation— in practical terms, the capacity to mine the data your projects already produce without hiring a knowledge-management team. This is the AI implementation services story most vendor marketing buries: tools are scale enablers, not replacements for architectural judgment.

The academic foundation for what these tools enable is older than the tools themselves. A peer-reviewed framework for lessons-learned databases describes the workflow in four phases¹⁷:

• Collection— capture lessons as they accumulate during the project
• Analysis— categorize and identify patterns across captures
• Implementation— feed conclusions into future projects
• Culture— leadership and resources that sustain the system

The AIA Firm Survey 2024 shows 61% of large firms use AI in day-to-day work versus 27% of small firms¹⁸— a gap that AI-enabled CA tools are positioned to close. Knowledge management is becoming a key organizational capability for competitive advantage in construction¹⁹, and small firms now have access to the same data-capture infrastructure that used to require a knowledge-management team.

One honest caveat: tools alone don't close the loop. A firm has to commit to reading what the tools surface. The capacity argument settles whether you can do this. The financial argument settles whether you should.

The ROI of Closing the Loop

The financial case for closing the design-construction feedback loop is straightforward. RFIs cost $1,080-$3,000 each when including indirect costs¹. 70% of change orders trace to design errors². And even modest reductions in recurring design weaknesses compound across every future project the firm runs.

If your typical project generates 200 RFIs and 70% of them trace to design gaps, a 10% reduction over time is roughly $20,000-$60,000 per project— before counting prevented change orders. Design feedback systems are the rare investment whose return compounds with every project the firm runs.

A rough illustrative example (not a documented case study):

A $40M project generates 200 RFIs at $2,500 each = $500K in cumulative response cost. A 10% reduction tied to systematized lessons learned = $50K back on that project alone. Multiply across the firm's annual project portfolio and the system pays for itself before counting reduced change orders or the 50% delay reduction Buildots reports on documented projects¹⁵.

Where the savings actually show up:

• RFI volume— fewer field questions because patterns informed the next set of drawings
• Change order frequency— design-driven rework drops as recurring gaps get addressed upstream
• Schedule— delay reduction compounds across project portfolios

The honest caveat: this ROI calculation requires the firm to actually act on the feedback, not just capture it. See our analysis on measuring AI success and ROI for how to track this beyond the tool deployment. Math is one thing. The honest question is where a firm actually starts, especially without a CA team to dedicate to this.

Where to Start — Different Playbooks for Different Firm Sizes

Closing the design-construction feedback loop starts with capturing the data from one project, not building a firm-wide system. Small firms (<10 employees) should pick a single high-RFI project and audit it manually before investing in tools. Mid-market firms ($20M-$100M revenue) should pair a tool (OpenSpace, Buildots, or Procore Copilot, depending on existing stack) with a simple lessons-learned template. Large firms should formalize the audit into the project closeout process.

Start with one project. Audit it. Then decide what infrastructure earns its keep. The cheapest version of this is a spreadsheet and a discipline. The scalable version is AI plus that same discipline. For founders weighing tool choices and process changes simultaneously, our AI decision framework for founders covers the prioritization sequence.

Search queries like "s and h construction and design" often lead firm leaders to peers who've systematized this— not because they bought better software, but because they built an AI culture in your practice that treats CA data as design intelligence, not paperwork.

Firms working through which combination of tools, templates, and process changes fits their practice often benefit from an outside set of eyes. Dan Cumberland Labs helps founder-led firms map AI-enabled feedback systems to their actual workflows. None of this requires a perfect system. It requires the firm to decide the data is worth looking at.

FAQ

What is construction administration?

Construction administration is the architect's formal services during construction, including site inspections, RFI review, submittal approvals, change order documentation, and certificate of substantial completion. It is defined in AIA Document B101 (Owner-Architect agreement)³ and AIA Document A201 (General Conditions of the Contract for Construction)⁴.

Why do RFIs happen so often?

Approximately 70% of RFIs result from design errors, incomplete drawings, or missing specifications². The remaining share stems from site conditions, utility conflicts, or value engineering changes during construction⁵.

How much does an RFI cost?

Individual RFI responses cost construction firms an average of $1,080, according to a 2013 Navigant Consulting study reviewing 1,362 projects¹. Current estimates including indirect costs of delay and rework range from $2,000 to $3,000 per RFI.

What is post-occupancy evaluation (POE)?

POE is an independent evaluation of a building's performance after occupancy, measuring energy efficiency, indoor environmental quality, occupant satisfaction, and functional success against the design intent. The U.S. GSA's Whole Building Design Guide describes it as providing architects with insights to improve future design outcomes¹⁰.

Why don't more firms conduct POE?

POE requires budget, time, and client agreement— and there is rarely a contractual hook to fund it. Only about 3% of architectural practices regularly conduct formal POE, and roughly 50% never do, according to UK-focused research⁹¹¹.

The Feedback Loop Has Always Been There

Construction administration has always been the highest-fidelity feedback your design process will ever receive. What has changed is whether you can afford to listen.

The feedback loop isn't a new thing. The infrastructure to actually use it is. Capturing the loop used to require staff most firms couldn't afford; AI-enabled capture has lowered the cost of systematization to a point where any firm size can start. Pick one project. Look at its RFI log. See what it tells you. That's the entire first step.

References

1. eSub, "The RFI and Its Cost to Construction Firms" (2024) — https://esub.com/blog/rfi-cost-construction-firm
2. Rhumbix, "Change Orders in Construction: The Definitive Guide" (2025) — https://www.rhumbix.com/blog/change-orders-construction-definitive-guide
3. American Institute of Architects, "Summary: B101–2017, Standard Form of Agreement Between Owner and Architect" (2017) — https://help.aiacontracts.com/hc/en-us/articles/1500010280541-Summary-B101-2017-Standard-Form-of-Agreement-Between-Owner-and-Architect
4. American Institute of Architects, "Summary: A201–2017, General Conditions of the Contract for Construction" (2017) — https://help.aiacontracts.com/hc/en-us/articles/1500010259162-Summary-A201-2017-General-Conditions-of-the-Contract-for-Construction
5. SubmittalLink, "RFI Meaning in Construction: Plain-English Guide" (2024) — https://www.submittallink.com/post/rfi-meaning-in-construction
6. Monograph, "Guide to Architectural Fees" (2025) — https://monograph.com/blog/guide-to-architectural-fees
7. American Institute of Architects, "The 2024 Firm Survey Report: The Business of Architecture" (2024) — https://www.aia.org/resource-center/aia-firm-survey-report-2024
8. Procore, "RFIs: A Contractor's Guide to Requests for Information" (2025) — https://www.procore.com/library/rfi-construction
9. Taylor & Francis Online, "Post-occupancy evaluation in architecture: experiences and perspectives from UK practice" (2017) — https://www.tandfonline.com/doi/full/10.1080/09613218.2017.1314692
10. U.S. General Services Administration, "Post-Occupancy Evaluations" Whole Building Design Guide (2024) — https://www.wbdg.org/resources/post-occupancy-evaluations
11. PlanMan, "Post-Occupancy Evaluations & UK Architects: an Idea for Growth?" (2023) — https://www.planman.app/blog/architecture/post-occupancy-evaluations/
12. HMC Architects, "The Role of the Post Occupancy Evaluation in Architecture" (2020) — https://hmcarchitects.com/news/the-role-of-the-post-occupancy-evaluation-in-architecture-2020-01-22/
13. ResearchGate, "The need to improve double-loop learning and design-construction feedback loops: A survey of industry practice" (2014) — https://www.researchgate.net/publication/263526088_The_need_to_improve_double-loop_learning_and_design-construction_feedback_loops_A_survey_of_industry_practice
14. OpenSpace, "360 Reality Capture Software for Construction" (2025) — https://www.openspace.ai/products/capture/
15. Buildots, "AI-Based Progress Tracking is Automating Construction Schedule Updates" (2025) — https://buildots.com/blog/automate-schedule-updates/
16. Procore, "The Rise of AI Co-Pilots in Construction" (2025) — https://www.procore.com/library/ai-co-pilots-construction
17. ScienceDirect, "A Lessons Learned Database Structure for Construction Companies" (2015) — https://www.sciencedirect.com/science/article/pii/S1877705815031719
18. American Institute of Architects, "The 2024 Firm Survey Report: AI Adoption Data" (2024) — https://www.aia.org/resource-center/aia-firm-survey-report-2024
19. Emerald Insight, "Addressing the knowledge management 'nightmare' for construction companies" (2019) — https://www.emerald.com/insight/content/doi/10.1108/CI-02-2019-0013/full/html