What Happens When The Index Takes Vacation

The Two Indexes

Every architecture firm prints a cover sheet at the front of every CD set called the drawing index. Every architecture firm also runs on a second architecture index— the one in a senior principal's head— that nobody printed.

The first one is G-001, the sheet list on the cover of the construction document set, organized by the CSI and National CAD Standard discipline letters: G for general, A for architectural, S for structural, and on down the deck.⁷ The second one is the person you Slack when a junior asks why the wall section on a 2019 project handled the vapor barrier the way it did. It's the partner who remembers which AHJ inspector pushed back on the parapet detail last spring. It's the studio director who knows which vendor showed up when you needed glazing in eight weeks.

This second index doesn't appear on G-001. It rarely appears in a job description. And it isn't the "architecture" of software systems or data structures. This is architecture in the licensed-practice sense, and the index is the senior who has been in the room for thirty years of decisions.

The architecture index has two meanings. One is the sheet list on G-001. The other is the senior person who knows where every past detail, code interpretation, and lesson learned lives— and that one isn't printed.

When the unprinted index takes vacation, retires, or gets sick, the firm finds out fast how expensive a human index actually is. Most days, the unprinted index is in the building. Some days it isn't.

What "Takes Vacation" Looks Like

When the senior who serves as the firm's de facto knowledge index is unavailable, junior staff stall on questions only that person can answer, QA reviews back up, coordination across disciplines slows, and decisions get re-litigated because no one else remembers why the last call went the way it did.

The scene is familiar. A coordinator chases the principal on Slack while a deadline slips. A junior hunts for the right wall section from a 2017 project and doesn't know which one to grab. A discipline lead waits on a code call that only the senior has the case-law-of-the-firm to make. The work continues. It just gets quieter and slower.

Three operational symptoms show up consistently:

• QA bottlenecks. Senior-level review backs up because only the senior knows the firm's standards on the live question.¹¹
• Cross-discipline coordination slows. The senior was the translator between structural, MEP, and architectural; without them in the room, teams misalign.
• Re-litigation. The firm spends partner-level judgment on a question it already answered three years ago, because nobody indexed the answer.

Great American Insurance, writing for design professionals on succession planning, names the pattern: during transitions, firms experience reduced senior-level review, inconsistent application of QA/QC procedures, and gaps in project-specific knowledge that can leave active projects vulnerable and complicate professional-liability claims handling.¹¹ The Association for Talent Development adds the harder part— the highest-risk knowledge category is the tacit kind that lives in many years of experience, and few firms are equipped to capture it before a retiree leaves.¹²

During transitions, firms experience reduced senior-level review and inconsistent QA/QC application— the operational signature of an index that's offline.

The vacation version is reversible. The retirement version is not.

The Math: What an Absent Architecture Index Costs

Industry-wide, construction professionals spend 35% of their time on non-optimal activities, including an average of 5.5 hours per week per person hunting down project data such as revised drawings.¹ The average RFI costs $1,080 to review and respond to.² And 22% of rework comes from inadequate project information.¹

These figures cover the whole construction industry. The architecture-firm slice rolls up inside them, and at the firm level the cost shows up as quiet margin erosion that no line item names. The mechanic is the same regardless: when the index lives in one head, the rest of the firm spends time hunting, asking, and waiting.

The 2018 PlanGrid + FMI study Construction Disconnected put the U.S. industry tab at $177.5 billion annually in labor cost from non-optimal time use, plus another $31 billion in rework— with 22% of that rework traced to inadequate project information.¹ Every hour a firm spends finding a past decision is an hour it isn't billing or improving the next decision. That's part of the hidden costs of AI projects too: the costs that compound without ever earning a line item.

The RFI math comes from the 2013 Navigant Construction Forum study, which analyzed 1.1 million RFIs across 1,362 projects.² Median response time: 9.7 days. Share that received no response at all: 21.9%.

The clock is the second pressure. The American Institute of Architects' 2023 demographics summary shows 13% of AIA-member architects are 65 or older and another 22% are between 55 and 64.⁴ Median member age is 51.⁵ The U.S. Bureau of Labor Statistics projects roughly 7,800 architect openings per year, many of them driven by retirement.⁶

Roughly 35% of AIA-member architects are 55 or older. The human index in most firms is retiring on a knowable curve.

If the cost is this large, why does every firm still let the index live in one head?

Why It Always Concentrates on One Person

Tacit knowledge transfers slowly. Documentation captures the what of a decision, rarely the why. As firms scale past the founder's personal capacity, the index quietly compounds in the same one or two heads it always lived in— and almost no firm re-engineers it on purpose.

Tacit knowledge transfers slowly. Documentation captures the what of a decision, rarely the why.

The Association for Talent Development frames it cleanly: in most organizations the highest knowledge-loss risk is the tacit kind that comes with many years of experience, and few firms are equipped to capture it before a senior leaves.¹² Starmind, writing about professional services specifically, observes that documents capture semantic content (the conclusion) and fall short on episodic knowledge (the trade-offs considered, the why behind the call).¹³

The traditional answer has been shadowing. Hire a junior, sit them next to the senior, let the transfer happen by osmosis over a decade. It works partially. The timing is the issue: tacit knowledge transfers slowly enough that the senior often retires before the shadow has absorbed the firm-specific case law.

Firms scale through hiring and offices, not through re-engineering the index. You buy Newforma to manage documents. You buy Deltek for project management. You build a detail library on a shared drive. The senior remains the synthesizer because that's where the synthesis has always lived. You can't read the label from inside the bottle— the pattern stays invisible until the index takes vacation.

Why Newforma and ACC Alone Aren't the Architecture Index

Newforma, Autodesk Construction Cloud, and SharePoint are storage and version-control layers. They organize what the firm already wrote down. They do not synthesize across a 30-year archive, and they do not answer the questions a junior actually asks the senior.

Document management does several things well: storage, versioning, project communication audit trail, transmittal tracking. These are the foundation, and most firms have already paid for them.

What document management does not do:

• Answer "show me a similar wall section from a coastal commercial project where the AHJ pushed back on the vapor barrier."
• Surface the reasoning behind a past detail choice when the architect of record has moved on.
• Connect a current question to the three projects where the firm already solved a version of it.

LS3P is the cleanest published case here. The firm runs 320 people across eight offices and five decades of project data. Its stack already includes Deltek Vision, Planifi, Newforma, BirdDog, and Open Asset. Even with that stack in place, LS3P created a dedicated Data Manager project role and a custom internal Dashboard to make the underlying data queryable for the practice.⁹ The stack was necessary. The synthesis layer on top was a separate build.

Storage holds. The architecture index synthesizes.

The 2018 PlanGrid/FMI research found another wrinkle: more than 75% of construction firms provide mobile devices to staff, yet fewer than 20% consistently use apps beyond email, text, and phone.¹ Buying the tool and solving the index problem are two different events.

What sits on top of the storage layer matters. That layer has a name now.

Making the Architecture Index a System

A queryable knowledge index— built on top of the firm's existing archive— turns 30 years of decisions into something a junior can ask in natural language. The technical name for this is RAG, retrieval-augmented generation: the AI retrieves real content from the firm's documents and uses it as the basis for an answer instead of inventing one.⁸

Christopher Parsons, founder of Knowledge Architecture, puts it directly: "RAG prevents hallucinations because it is grounded in real data."⁸ The model pulls from the firm's own projects, code interpretations, and lessons learned.

Several AEC firms are already running production versions:

• Shepley Bulfinch made 400+ internal videos searchable with AI search and transcription. As VP/CIO Jim Martin describes it, the system summarizes spoken content and links directly to the exact spot in the video where a topic is discussed.⁸
• MBH Architects uses AI search to generate proposal responses from past projects and technical expertise in minutes— a task the marketing team previously spent hours on, gathering content manually.⁸
• BORA Architecture and Interiors made recurring sustainability questions searchable. Director Corey Squire: "There are certain questions that are asked again and again… now, everything is documented and searchable."⁸
• Ware Malcomb reduced feasibility studies from 3 days to a few hours, saving an estimated $1,200–$1,350 per study and more than $200,000 annually.¹⁰

These are 100+ person firms, and the unit economics scale down. A $20M–$100M firm can start narrow with the same playbook.

What to index first, in order:

1. Proposal precedent. Highest near-term ROI. Pulls from existing wins.
2. Recurring code interpretations and AHJ-specific calls. The questions that come up every year.
3. Detail libraries. Wall sections, head/sill/jamb, edge conditions— the recurring construction-document moves.
4. Lessons learned and post-occupancy findings. The expensive knowledge the firm paid for once and shouldn't pay for again.

Tools come second. Deciding what's worth indexing is the thinking that has to happen before any tool decision.

The priority list comes first; the implementation follows— and that's where building AI culture inside a firm and measuring AI success become the operational guardrails.

The adoption curve is moving fast enough to matter. Deltek's 2025 Clarity study reports 53% of A&E firms now use AI tools, up from 38% the prior year, with 44% planning to increase AI investment in 2025— the highest of any technology category— and 87% confident AI will help them offer more services.³ At the individual-architect level, adoption is still concentrated in larger firms: Monograph's December 2025 analysis found only about 6% of architects routinely use AI today.¹⁰ The gap between firm-level activity and individual-practitioner routine is exactly the window a mid-market firm can move through.

None of this replaces the senior. Here's why that matters.

What Stays Human

AI surfaces and synthesizes. The licensed professional stamps the set, owns the code interpretation in jurisdiction, carries the liability, and runs the call with the client. None of that moves to a model.

What AI does well in this context: retrieval, summarization, surfacing precedent, drafting first-pass language, transcribing and indexing meetings, generating proposal content, answering "where have we done X before?"

What stays with the licensed professional:

This is the operational point. The goal is to free the senior from being the lookup table, so the senior's time goes to the calls, the stamps, and the work AI structurally cannot do. AI amplifies the senior's judgment. The stamp stays with the senior.

Great American Insurance frames the underlying risk: QA/QC inconsistency and gaps in project-specific knowledge during transitions are exactly what creates professional-liability exposure.¹¹ A queryable knowledge index reduces that exposure because the firm's accumulated judgment stops being one-person-deep.

AI does the first pass. The licensed professional does the final one. Both are required.

AI amplifies human capability. No matter the question, people are the answer.

Where to Start

If the index in your firm has a name and a calendar, the first move is deciding what's worth indexing— and that decision belongs to the senior whose head currently holds it.

A pragmatic sequence:

1. Interview the senior(s) for what they actually get asked.
2. Start narrow with proposal precedent or recurring code Q&A.
3. Put a retrieval layer on top of what's already in storage.
4. Iterate.

The senior makes the indexing call. The system catches what one head can't hold.

The drawing index goes on G-001. The architecture index— the one that runs the firm— belongs on a system the firm can query.

Principals who'd rather not piece this together solo can bring a partner who's mapped this transition before. Dan Cumberland Labs helps founder-led AEC firms move from a person-shaped index to a queryable system— through our AI implementation services— without overclaiming what AI can do in licensed practice. If that's the work, an AI decision framework for founders is a useful companion read before any conversation.

FAQ

A few of the questions that come up most often when firms read this:

What is a drawing index in architecture?

The drawing index (also called the sheet index or sheet list) is the list of every sheet in a construction document set, typically located on the G-001 cover sheet at the front of the set. It's organized by CSI and National CAD Standard discipline designators: G for general, A for architectural, S for structural, and on down the deck.⁷

How much does an RFI cost?

A 2013 Navigant Construction Forum study analyzing 1.1 million RFIs across 1,362 projects found the average cost to review and respond to a single RFI was $1,080, with a median response time of 9.7 days.² Roughly 21.9% of RFIs received no response at all.²

How much time do construction professionals spend looking for project data?

According to the 2018 PlanGrid + FMI Corporation Construction Disconnected report, construction professionals spend 35% of their time on non-optimal activities, including 5.5 hours per week per person hunting for project data such as revised drawings.¹ The industry loses $177.5 billion annually to non-optimal time use.¹

What percentage of architects are near retirement?

According to the AIA's 2023 Membership Demographics Report, 13% of AIA-member architects are 65 or older and another 22% are between 55 and 64.⁴ The median member age is 51.⁵ Roughly one in three AIA-member architects is within 10 years of standard retirement.

What is RAG and why does it matter for an architecture firm?

RAG (retrieval-augmented generation) is an AI architecture that retrieves relevant content from a knowledge base— such as a firm's project archive— and uses it as the basis for a generated answer. As Knowledge Architecture founder Christopher Parsons puts it, "RAG prevents hallucinations because it is grounded in real data."⁸ For an architecture firm, RAG is the technical layer that turns the existing archive into something a junior can query in natural language.

Does AI replace the senior architect?

No. AI surfaces and synthesizes content from a firm's archive. The licensed professional still owns code interpretation in jurisdiction, stamps the set, runs the final QA/QC, and carries the liability. AI does the first pass; the licensed professional does the final one.¹¹

References

1. PlanGrid + FMI Corporation, "New Research from PlanGrid and FMI Identifies Factors Costing the Construction Industry More Than $177 Billion Annually" (2018)— https://www.prnewswire.com/news-releases/new-research-from-plangrid-and-fmi-identifies-factors-costing-the-construction-industry-more-than-177-billion-annually-300689826.html
2. Navigant Construction Forum (Hughes, Wells, Nutter, Zack), "Impact & Control of RFIs on Construction Projects" (2013)— https://www.cmaanet.org/sites/default/files/resource/Impact%20&%20Control%20of%20RFIs%20on%20Construction%20Projects.pdf
3. Deltek, "46th Annual Deltek Clarity Architecture & Engineering Industry Study" (2025)— https://www.prnewswire.com/news-releases/ai-talent-and-record-profits-what-the-46th-annual-deltek-clarity-ae-study-reveals-about-the-architecture--engineering-industry-302453217.html
4. American Institute of Architects (via Architect Magazine), "Shifting Gears in Today's Workforce" (2024)— https://www.architectmagazine.com/aia-architect/aiafeature/shifting-gears-in-todays-workforce_o
5. American Institute of Architects, "AIA Membership Demographics Report 2023" (2024)— https://www.aia.org/sites/default/files/2024-09/AIA_Demographics_Report_2023.pdf
6. U.S. Bureau of Labor Statistics, "Architects: Occupational Outlook Handbook" (2024)— https://www.bls.gov/ooh/architecture-and-engineering/architects.htm
7. Archtoolbox, "Construction Document Sheet Numbers and Order"— https://www.archtoolbox.com/construction-document-sheet-numbers/
8. Friedman & Partners, "The Rise of AI Search in AEC Knowledge Management" (2025)— https://friedmanpartners.com/the-rise-of-ai-search-in-aec-knowledge-management/
9. LS3P, "Knowledge Management: Big Data in Service to the Business and Practice of Architecture"— https://www.ls3p.com/knowledge-management-big-data-in-service-to-the-business-and-practice-of-architecture/
10. Monograph, "Artificial Intelligence Architecture: Use Cases & Adoption" (2025)— https://monograph.com/blog/artificial-intelligence-architecture-use-cases-adoption
11. Great American Insurance Group, "Succession Planning: A Critical Risk Management Issue for Design Professionals"— https://www.greatamericaninsurancegroup.com/content-hub/news-details/succession-planning--a-critical-risk-management-issue-for-design-professionals
12. Association for Talent Development, "6 Steps to Take Before Key Employees Retire"— https://www.td.org/insights/6-steps-to-take-before-key-employees-retire
13. Starmind, "When Your Firm's Most Valuable Tacit Knowledge Walks Out the Door" (2026)— https://www.starmind.com/blog/most-valuable-tacit-knowledge