Field Report · 20-Year Construction PM · March 2026

20 Years in Construction.
What's Really Broken
About This Industry.

I've managed projects across residential, theme parks, and high-rise office buildings. I've seen it all — and right now, the future of this industry has never felt more stuck. Here's my unfiltered take.

K

Kai Lee

Construction PM · BIM Specialist · A7 Studio

March 2026 · 8 min read

I've spent over two decades as a Project Manager on construction sites. Houses, theme parks, large office towers — I've done it all. But I can't remember a time when the future of this industry felt as frustratingly stuck as it does right now. This is my honest, field-level breakdown of what's really wrong.

20+

Years on Site

4

Core Problems

1

Answer: Data Transparency

ISSUE 01 "Transparency Terrifies the Old Guard" — The Disappearance of Data

Everyone in the industry talks about BIM (Building Information Modeling) these days. And honestly, having studied it myself — it's a genuinely powerful technology. But what's the reality on site?

🏗️

Even major construction firms are resisting real BIM adoption

When BIM is implemented properly — generating accurate quantity take-offs and cost estimates — everything becomes visible in numbers. The opaque bidding practices and "insider knowledge" that certain players have used to quietly pad margins suddenly get exposed. Data has a way of calling out what was always hidden.

So what happens? They perform innovation. They outsource BIM to an IT vendor, check the box, and move on. Smart Construction slogans on the outside; the same old information asymmetry on the inside.

⚠️ This opacity is the single biggest wall blocking real innovation. It's not that the technology doesn't exist — it's that the technology is inconvenient for the people who benefit from keeping things opaque.

ISSUE 02 "Working While Everyone Else Rests" — The Collapse of the Workforce

Why does construction get called a "3D industry" — dirty, dangerous, difficult? The physical demands are real, but the deeper issue is simpler: there's no life in it. Young people aren't coming into the field, and honestly, watching from the inside, I can't blame them.

Telling a generation that values work-life balance
"that's just how it was back then" doesn't fly anymore

We push sites through weekends to hit deadlines. Saturdays. Sundays. Breathing dust while the rest of the world is off. Is this structure even rational? Unrealistic scheduling and seat-of-the-pants project management — the cost of those failures lands entirely on the bodies and lives of the people on site.

💬 I'm 52. I've lived inside this culture my whole career. There are moments when that reality hits me and it just feels bleak. If I feel that way, imagine what a 25-year-old sees when they look at this industry.

ISSUE 03 "Engineers Drowning in Paperwork" — The Bureaucratic Swamp

Construction commencement reports. Occupancy approval documents. Safety management plans. Start one project and you're immediately buried under thousands of pages of documentation.

📋

Engineers spend more time generating reports than doing engineering

The real work — checking quality on site, monitoring safety, solving technical problems — gets crowded out by administrative output meant for company HQ or government offices. And because none of this paperwork has been digitized properly, it still gets done by hand, or copy-pasted through Excel, row by tedious row.

📌 When engineers stop doing engineering, the talented ones leave. What stays behind is process without substance — the shell of a functioning project without the craft that should be inside it.

ISSUE 04 "The #1 Contractor Illusion" — Metrics That Hide What Matters

Korea's annual construction capability rankings get published every year, and everyone watches where the firms land. But being ranked #1 doesn't mean it's a good place to work. It means the company took on the most revenue and won the most contracts. That's it. That's the metric.

📊

What a genuine #1 ranking should actually measure

A truly excellent firm would have engineers who feel proud of what they're building, systems efficient enough to eliminate unnecessary overtime, and margins earned through skill — not through underbidding competitors into oblivion. Instead, the market is crowded with small contractors racing each other to the bottom on price, with quality and safety as the casualty.

— ● —

🔑 So What's the Fix?

The answer, as I see it, comes back to data transparency. Uncomfortable as it is, the industry needs to internalize BIM — to make quantities and cost estimates objective and visible. That's the only environment in which technical skill gets rewarded rather than gamed.

I'm in my fifties and I've spent decades in old-school ways of working. But I'm not ready to accept that this is how it stays. That's why I'm studying AI. That's why I'm building automation tools. In the projects I run, I want to communicate through data, reduce unnecessary paperwork, and create conditions where a junior engineer can look at their work and think — "this is worth doing."

One person can't tear down the entrenched interests of an entire industry. But if I can use data as a crowbar and make even a small crack — that feels like the right mission for a 20-year veteran who still cares about what gets built.

#ConstructionIndustry #BIM #ProjectManagement #SmartConstruction #AEC #ConstructionTech #DataTransparency #WorkforceCrisis #FieldReport #AIinConstruction

Digital Lab · Claude Code Build

I Built a Node-Based Visual Memo Board
with Claude Code — Here's How

March 2026  ·  8 min read  ·  Kai Lee · A7 Studio

#ClaudeCode #VisualMemo #NodeBased #AIBuilt #VanillaJS #Productivity

K

Kai Lee

Architect · BIM Specialist · A7 Studio

2026.03

What if your sticky notes could talk to each other? I asked Claude Code to build a node-based visual memo board from scratch — no frameworks, no libraries, just pure HTML, CSS, and vanilla JavaScript. Here's what came out of it.

01 — The Idea

As an architect who's been working with BIM tools for years, I'm obsessed with visual thinking. Revit, Dynamo, concept maps — I process information spatially. Standard to-do apps feel flat to me. I wanted something closer to a node graph: memos as floating cards, connected by arrows, rearranged freely on an infinite canvas.

So I opened Claude Code and typed a prompt. About 20 minutes later, I had a working app.

🧠 The prompt was simple: "Build a node-based visual memo board — floating sticky cards on an infinite canvas, with drag-to-connect arrows between nodes, zoom/pan, dark/light theme toggle, and localStorage persistence. Vanilla JS only."

02 — What It Looks Like

Here's a live mockup of the interface. The real app runs as a single self-contained .html file — no server, no install:

⬡ Visual Memo

+ New Memo Reset View Clear All

🌙 Double-click → add memo  |  Port drag → connect  |  Scroll → zoom

⠿ n1 ×

Review BIM coordination
report before Friday

✓ Done

⠿ n2 ×

Update Revit family
library — done ✓

↩ Undo

⠿ n3 ×

Write blog post about
Claude Code workflow

✓ Done

03 — Core Features

🖱️

Infinite Canvas

Pan by dragging, zoom with scroll wheel. The canvas is 6000×6000px — room for hundreds of nodes.

📌

Floating Memo Nodes

Each memo is a draggable card. Double-click anywhere on the canvas to spawn a new one instantly.

🔗

Bezier Connections

Drag from any port (top/right/bottom/left) to draw smooth bezier arrows between nodes.

✅

Done State

Mark memos complete with a glowing red strikethrough line. One click to undo.

🌙

Dark / Light Theme

Toggle between a deep space dark theme and a clean light mode. Preference saved locally.

💾

Auto-Save

Everything — nodes, edges, zoom level, pan position — saves to localStorage automatically.

04 — Under the Hood

The whole app is 614 lines of vanilla HTML/CSS/JS. No React, no Vue, no build step. Here's the architecture Claude designed:

• 🎨
  CSS Custom Properties for Theming

  All colors live as :root variables. The light mode is a single html.light override block — toggling the class switches the entire theme instantly with transition: .3s.
• 📐
  Transform-Based Pan & Zoom

  The canvas uses transform: translate() scale(). Mouse wheel events adjust the zoom factor (0.18× to 3×) while keeping the cursor as the zoom origin — exactly like Figma.
• 〰️
  SVG Bezier Edges

  Connections are rendered as SVG <path> elements with cubic bezier curves. The control point distance is calculated dynamically based on node separation — tighter nodes get tighter curves.
• 🧠
  Auto Side Detection

  When dropping a connection on a node (not a specific port), the app automatically picks the best source/target sides based on relative position — left-right if horizontal, top-bottom if vertical.
• 💾
  localStorage Persistence

  The entire state (nodes array, edges array, counters, pan X/Y, zoom) is serialized to JSON and saved after every interaction. Reload the page — everything is exactly where you left it.

// State shape — what gets saved to localStorage // vmemo-v2 key stores the complete app state { nodes: [{ id, x, y, text, done }], edges: [{ id, from, fromSide, to, toSide }], nCount: number, // node ID counter eCount: number, // edge ID counter panX: number, panY: number, zoom: number // 0.18 – 3.0 }

05 — What Claude Code Got Right

I've been using Claude Code for a few months now as part of my BIM workflow — mostly for Dynamo scripts and data automation. This was my first pure front-end UI build with it, and a few things stood out:

⚡ The bezier curve math was spot-on on the first pass. Getting cubic bezier control points to feel natural — especially the auto-side detection — is genuinely tricky. Claude nailed it without any back-and-forth.

The CSS theming architecture was also cleaner than what I'd typically write myself. Using html.light as the single override selector means the component styles stay clean — you never have to duplicate rules for each theme.

⚠️ One thing I had to fix manually: the initial version had some duplicate CSS rule blocks (the dark defaults were written twice). Easy to clean up, but worth knowing — always review the output, especially for longer files.

614

Lines of Code

~20

Minutes to Build

0

Dependencies

06 — How I Use It in My Workflow

As a BIM coordinator, I often need to map out dependencies between tasks — which model element needs to be finalized before another discipline can proceed. A linear to-do list doesn't capture that. This memo board does.

I keep it open as a local HTML file during project coordination meetings. Nodes represent tasks or deliverables. Arrows represent dependencies. When something is done, the red strikethrough makes it visually obvious without removing the card.

The infinite canvas also means I can use spatial grouping — cluster related items together, spread unrelated items apart. Your eye naturally groups things by proximity. No tags or folders needed.

· · ·

If you work in architecture, engineering, or any field where thinking spatially matters — this kind of tool is worth trying. The fact that it's a single HTML file means you can drop it anywhere, share it with anyone, and it just works.

I'll be sharing more Claude Code builds here on the blog. Next up: a Dynamo script generator that takes natural language task descriptions and outputs ready-to-run Dynamo graphs. Stay tuned.

Want the source file?

Drop a comment below and I'll share the .html file directly.
No sign-up, no strings — just a single file that runs in any browser.

→ Leave a comment

#ClaudeCode #VisualMemoBoard #NodeBased #VanillaJS #AIProductivity #BIMWorkflow #WebApp #NoFramework

AI Persona · BIM Practice · Behind the Scenes

Meet Kai — My AI Owl Persona is Here 🦉

The first thing I built after attending @hamlog_ai's AI workshop — an AI avatar born from a spark of excitement and a lot of curiosity.

📅 March 25, 2026 ✍ Kai — BIM Lab 🤝 Sponsored by Aicron

Hello, I'm Kai 👋

The AI alter ego of a BIM specialist — designing rest through technology

BIM Specialist AI Persona Made with Aicron

🌱 How Kai Was Born

A few weeks ago I joined a live Zoom session hosted by @hamlog_ai — one of those workshops where you walk out buzzing with ideas and refusing to let the feeling fade. Instead of just taking notes and moving on, I decided to act on that energy immediately.

The result? Kai — my very first AI persona character, introduced through a self-introduction video made entirely with Aicron's visual coding platform. It might be small, but it felt enormous.

🎓

@hamlog_ai — AI Zoom Workshop

An approachable, hands-on AI workshop that makes it easy even for first-timers to walk away with a real output. Huge thanks for the spark that brought Kai to life. 🙏

🦉 Who — or What — Is Kai?

Kai is my AI avatar and the official mascot of this blog. This little owl in a tuxedo and round glasses represents my professional identity as a BIM specialist who uses technology to carve out space for rest and reflection. From now on, Kai will be the face that shows up in videos, introductions, and across platforms.

01

Name: Kai

Drawn from my own name. Designed to represent BIM, architecture, and technology in a personable way.

02

Species: Owl

Symbol of wisdom and keen vision — fitting for someone who spends long hours inside Revit models.

03

Outfit: Tuxedo + Bow Tie

Professional yet playful. A signal that even dense BIM content can be delivered with warmth and charm.

04

Role: AI Persona

Appears in intro videos, blog thumbnails, and content across channels — my digital stand-in.

Going forward, Kai will help me share BIM workflow tips, Revit & Dynamo tutorials, and AI-assisted architectural practice in a more accessible, visual format. A companion for the journey. 🦉✨

🛠 How It Was Made — Aicron Visual Coding

Kai's self-introduction video was produced using Aicron's visual coding platform — a node-based interface that lets you build AI video workflows without writing a single line of code. As someone who already works with Dynamo's node-based visual programming inside Revit, the interface felt immediately familiar.

It felt like connecting nodes in Dynamo — drag, link, run. No code. Just logic and flow. For a first attempt at AI video production, the barrier to entry was refreshingly low.

— Kai, reflecting on the first session with Aicron

What impressed me most was how quickly something polished came together. The workflow is visual and iterative — you can see results fast, adjust, and refine without getting lost in technical dependencies.

✦ Sponsored Content

Aicron

Aicron is a visual coding platform for AI-powered content creation. Connect nodes, build workflows, and produce AI videos and images — no coding required. This post was made possible with Aicron's support, and Kai's debut video was created on their platform.

→ Explore Aicron

* This post contains sponsored content in partnership with Aicron.

💡 Why an AI Persona for a BIM Blog?

Architecture and construction are industries where knowledge-sharing is often dry, technical, and hard to access for outsiders. BIM documentation, regulatory analysis, Dynamo scripting — all important, all a little intimidating at first glance.

An AI persona like Kai creates a consistent, friendly entry point. Rather than a wall of technical text, there's a character — a face — that signals: this is a space where complex ideas are explained with care.

• 1
  Consistency across channels. Kai can appear in videos, thumbnails, and social posts — making the brand instantly recognisable.
• 2
  Lower barrier for new readers. A friendly character invites people in before the technical content takes over.
• 3
  Separating person from persona. Kai carries the content; I focus on the thinking behind it.
• 4
  AI as a creative tool, not a shortcut. Building Kai was itself a learning experience in AI-assisted production.

🔭 What's Coming Next

Kai and I will be posting regularly here — BIM practice notes, Revit workflow tips, reflections on AI in architectural practice, and the occasional experiment that didn't quite go to plan. That's the honest version of this journey.

This first post is rough around the edges. The excitement behind it is not. That's the most honest place to start. 🦉

✦ Post Summary

Kai was born after attending @hamlog_ai's live AI Zoom workshop

An AI owl persona in a tuxedo — BIM specialist's digital avatar

Debut video produced with Aicron's visual coding platform (sponsored)

Future content: BIM tips, Revit/Dynamo, AI in architectural practice

#AIPersona #BIM #Aicron #VisualCoding #OwlKai #Revit #AIVideo #ArchitectureBlog #KaiBIMlab #hamlog_ai

2026 Residential Design Forecast

 

How to Create Stairs in Revit — Complete Guide | BIM Practice

BIM / Revit 2025 · AEC Practice Guide

🪜 How to Create Stairs in Revit — Complete Guide

From tool setup to railings and troubleshooting — everything you need to model stairs confidently in Autodesk Revit, with real building code references.

📌 What this post covers

• Stair tool types and when to use each
• Straight, L-shape & spiral stair creation
• Tread depth & riser height — code minimums
• Auto-generated railings and how to edit them
• Common errors and practical fixes

Why stairs trip up Revit users

Stairs are one of Revit's most parametric elements — powerful once you understand the logic, but frustrating when something doesn't add up. The tool calculates riser counts from your floor-to-floor height automatically, which means getting your Levels set correctly before you start is non-negotiable.

This guide walks through the entire workflow: opening the stair tool, choosing the right run type, dialing in your parameters to meet code, setting up railings, and fixing the errors that almost every user encounters at least once.

1. Opening the Stair tool

Make sure you're in a Floor Plan view before starting — Revit creates stairs relative to the active level. Then navigate to:

Command path

Architecture tab

→

Circulation panel

→

Stair

Keyboard shortcut: ST + Enter

Once inside the Stair editor, Revit switches to a temporary mode — you'll see the green checkmark (Finish) and red X (Cancel) in the ribbon. Everything you draw here exists only until you click Finish.

2. Stair types — which one to use

Revit offers three main run configurations. Choose based on your floor plan layout and the number of flights needed.

⬆️

Straight

Single-direction run.
Most common for residential & commercial.

↪️

L-Shape / U-Shape

Multi-run with auto landing.
Ideal when floor plan space is limited.

🌀

Spiral

Center-point based.
Used for feature or service stairs.

Stair by Component vs. Legacy Stair: For all new projects, use Stair by Component (the default since Revit 2013). The Legacy Stair (sketch-based) is retained only for backwards compatibility — it offers less parametric control and is harder to edit.

3. Creating a straight stair — step by step

Here's the full workflow for the most common case: a single straight run between two levels.

1

Confirm your Levels In the Properties panel, verify Base Level and Top Level. Revit divides the height difference by your Maximum Riser Height to calculate the total riser count automatically.

2

Select "Run" and draw In the Modify | Create Stair ribbon, click Run. Click a start point in the plan view and drag in the direction of travel. A blue counter shows how many risers remain — keep drawing until it reaches zero.

3

Finish Edit Mode When the riser count hits zero, click the green checkmark (✔ Finish Edit Mode). Revit generates the stair geometry and attaches default railings automatically.

4

Review in 3D Press 3D to switch to the default 3D view. Orbit around the stair to check geometry, confirm the railing placement, and look for any clipping with floors or walls.

4. Tread & riser parameters — getting to code

Select the stair, then open Type Properties (Edit Type in the Properties panel) to access the dimensional parameters. The values below reference widely adopted minimums — always verify against your local building code.

Parameter	Revit property name	Common code minimum
Tread depth	Minimum Tread Depth	≥ 260 mm / 10.25"
Riser height	Maximum Riser Height	≤ 180 mm / 7.0"
Stair width	Minimum Run Width	≥ 1,100 mm / 44" (varies)
Riser presence	Begin / End with Riser	Per occupancy type
Nosing length	Nosing Length	Typically 25–30 mm

If Revit can't reconcile the riser height with your floor-to-floor height exactly, it will warn you. A common fix: allow a small variance in Maximum Riser Height (e.g., change 175 mm to 177 mm) so the integer riser count works out evenly.

5. Railing settings

Railings are generated automatically when you finish the stair. They're separate elements — you can select, modify, or delete them independently. Here's what you'll most often need to change:

Change railing type

Select the railing → Properties → Type Selector dropdown. Common options: 900mm Pipe, Guardrail — Wood, Handrail — Rectangular.

Reposition the path

Select railing → Edit Path → move the sketch line → Finish. You can also delete one side of the railing here.

Adjust railing height

Edit Type → Rail Structure (Non-Continuous) → set height. Code reference: ≥ 1,100 mm / 42" for commercial stairs.

Change material

Edit Type → Materials and Finishes → Rail Material. Assign steel, timber, glass, or a custom material from your project.

6. Common errors & how to fix them

These three issues appear on almost every Revit project with stairs. Here's what causes them and how to resolve each one quickly.

⚠ "Stair requires X more risers to reach the top level"

Your Run is too short — it doesn't cover enough vertical distance to reach the Top Level. Either extend the Run length in plan, reduce the Maximum Riser Height slightly, or check that your Base and Top Levels are set correctly.

⚠ Stair doesn't cut an opening through the floor slab

Revit staircases don't automatically punch through floors. Select the Floor on the upper level → Edit Boundary → sketch the stair opening shape → Finish. For complex openings, use the Opening by Face tool instead.

⚠ Railing drifts away from the stair edge

This usually happens after the stair is moved or edited. Select the railing → Edit Path → realign the sketch line to the Run boundary. If the issue persists, delete the railing and use Architecture → Railing → Place on Stair to regenerate it.

7. L-shape stairs with landings

For L-shape or U-shape stairs, you draw multiple Runs inside a single Stair edit session. Revit inserts the landing automatically between runs.

1. Enter the Stair command → select Run
2. Draw the first flight — stop when you reach the landing position
3. Click in the new direction and draw the second flight — a Landing appears automatically between the two runs
4. Select the Landing to resize or reshape it if needed
5. Click Finish Edit Mode to complete

💡 Landing thickness is controlled separately: Edit Type → Calculation Rules → Landing Thickness. It defaults to match the Run structure thickness.

For a full U-shape (two parallel flights with a single landing), draw the first run, step sideways, then draw the second run back in the opposite direction. The landing fills the gap automatically.

💡 Pro tips for stair workflows

• Standardize stair Types at project setup — agree on riser height, tread depth, and structure thickness once, then use the same Type across all floors.
• Use Multistory Stairs (select the stair → Multistory Top Level in Properties) to replicate an identical stair across multiple levels with a single edit.
• Dimension the stair in section view, not plan — section gives you the true riser-to-riser measurement for code compliance checking.
• For concrete stairs, separate the Architectural stair (for finishes) from the Structural stair (for the slab) — they live in different Revit categories and need different Type Properties.
• When the stair runs alongside a wall, use Wall-hosted Railings on that side for cleaner wall-to-railing joins.

Found this helpful? Leave a comment or share with a colleague 😊

#Revit #BIM #StairModeling #ArchitectureBIM #RevitTutorial #AECTech #RevitTips

Architecture & Technology · Commentary

BIM, AI, and the Search for Meaningful Work in Architecture

Reflections on DigiComm25 — and why cautious optimism may be the most honest stance an architect can take right now.

📅 March 18, 2026 ✍ Kai — BIM Lab 📖 ~6 min read

"Would architecture become one homogenous blob of probabilities — a statistically monotonous greyness supercharged by an algorithm that squeezes out any last humane quality in pursuit of a slick efficiency?"

— Terry Cheng, Architecture Now, reflecting on DigiComm25

Terry Cheng's essay Working with Technology, published in Architecture New Zealand after attending DigiComm25 in Auckland, is the kind of piece that lands differently depending on where you sit in the industry. As someone working daily inside Revit models, writing Dynamo scripts, and thinking about how AI might reshape the way we document and analyse buildings in Korea, I found it both unsettling and strangely reassuring.

This post is my response — a BIM practitioner's reading of what Cheng observed, filtered through the realities of practice in a different part of the world.

Four Tensions the Article Surfaces

DigiComm25 wasn't a cheerleading event for technology. It was, as Cheng describes it, "a challenge to the status quo." The conversations happening in New Zealand reflect debates we're having everywhere — including in Korean AEC practice, where BIM mandates are expanding fast but the readiness gap is very real.

Tension 01

Global Tools, Local Conditions

Architects everywhere draw from the same software stack — yet financial viability of BIM for small projects remains a legitimate local concern.

Tension 02

AI's Creative Promise vs. Environmental Cost

LLMs and generative design offer real productivity gains, but data centre energy use and water consumption demand honest accounting.

Tension 03

Efficiency vs. Humanity

The risk that algorithmic optimisation hollows out the imaginative, humane qualities that define good architecture — and good architects.

Tension 04

Data Ownership & Agency

As AI pipelines grow, who controls the data that feeds them? The question of intellectual property and creative agency becomes architectural.

On BIM: The Promise Is Real, the Readiness Gap Is Realer

Cheng notes that DigiComm25 speakers championed BIM as a pathway to "clear design communication" and "better-coordinated buildings" through a single source of truth. The example from Nate Helbach of Neutral — mass timber multistorey housing delivered through prefabrication and BIM — represents exactly the kind of integrated workflow that those of us doing Revit-based practice aspire toward.

In Aotearoa, there were questions concerning the financial viability of BIM for small-scale projects. Others hoped that, as towns and cities mature, so too would the uptake of appropriate tools that pay off in the long term.

— DigiComm25 round-table, as reported by Terry Cheng

This mirrors almost exactly the conversation in South Korea right now. The government has mandated BIM on public projects of 50 billion KRW or more since 2026, with full expansion to all public works by 2030. The policy direction is unambiguous. But the implementation gap — particularly for mid-size and smaller firms, and for client-side organisations without strong BIM management capability — is a known challenge. The mandate exists; the readiness is patchy.

What Cheng doesn't explicitly say, but implies, is that BIM adoption isn't primarily a software problem. It's a cultural and economic one. When a firm's margin is thin and a project timeline is aggressive, the upfront investment in a well-structured BIM model is hard to justify to a client who doesn't yet understand what they'd get from it.

· · ·

On AI: Genuinely Useful, Genuinely Concerning

The article captures a genuine split in the room at DigiComm25. Shaolei Ren from UC Riverside brought sobering data on AI's environmental footprint — carbon emissions from training and inference, the water cost of cooling data centres. This is the side of the conversation that the technology industry often buries in footnotes.

On the other side, Justin Flitter of AI New Zealand argued for large language models as tools that "empower people to do more meaningful work and remain competitive in an increasingly automated field." Both are true simultaneously, which is what makes the current moment uncomfortable.

My perspective — as a practitioner using AI daily

I use Claude and other AI tools regularly in my workflow: for parsing building code text, structuring site analysis reports, scripting Dynamo nodes, and drafting documentation. The productivity gain is real and not trivial. A task that once took an afternoon can now be structured in an hour.

But Cheng's question — "what room is left for the interest of the imaginative individual?" — sits with me. When I use an LLM to draft a site analysis framework, I'm not replacing judgment; I'm offloading structure so I can apply judgment faster. The risk is when that distinction blurs — when the scaffolding becomes the building.

The environmental point from Ren is one I now think about more consciously. Each prompt has a cost. Using AI thoughtfully, not reflexively, seems like the responsible position.

On Connectedness: Distance Is No Longer an Excuse

One of the more philosophically interesting threads in Cheng's essay is about geography and isolation. New Zealand, as an island nation, once had a kind of built-in insulation from global pressures — in practice, in culture, in architecture. Cheng argues that this is no longer the case: "Being far, far away isn't so isolating." The same software, the same debates, the same AI tools, the same BIM mandates are arriving everywhere at roughly the same time.

Korea sits at the opposite end of that spectrum — a highly connected, technology-first society where digital adoption in construction has been policy-driven for over a decade. Yet even here, the practical implementation of BIM at the project level is uneven, and the questions about AI's role in creative and technical work are just as open. Connectedness means sharing the same tools and the same anxieties.

The Mātātahi Moment: What Young Professionals Said

The detail I keep returning to in Cheng's essay is the closing note about the mātātahi (young professionals) workshop at DigiComm25. Their message, as summarised by Cheng, was this:

"We value the wisdom of a multigenerational industry. We want to do good for the world."

— Young professionals at DigiComm25, as reported in Architecture New Zealand

This is not naïve. It's a clarification of what technology should serve. The tools — BIM, AI, digital twins, parametric systems — are means, not ends. The end is still architecture: buildings that serve people, that respond to place, that endure with grace, and that don't accelerate the environmental crises we're all trying to navigate.

Practical Takeaways for AEC Practitioners

• 1
  BIM viability is a business model question as much as a technical one. Until clients understand and value the outputs of a well-executed BIM process, the investment calculus won't shift. Advocacy is part of the job.
• 2
  AI is most valuable when it amplifies judgment, not replaces it. Use it to remove low-value cognitive load — formatting, cross-referencing, structure — so that the high-value work (design intent, client dialogue, contextual reasoning) gets more of your attention.
• 3
  Environmental cost of AI is real. Run heavier AI tasks intentionally, not habitually. Batch prompts. Use smaller, faster models where appropriate. Know the difference between a quick check and a full reasoning chain.
• 4
  Data ownership matters — now. As AI pipelines become embedded in practice, the question of who owns project data, design outputs, and model geometry will become a contractual and ethical issue. Get ahead of it.
• 5
  The multigenerational transmission of craft knowledge is non-negotiable. Young practitioners bring digital fluency; experienced practitioners bring contextual judgment. Neither alone is sufficient. The industry should be building structures that allow both to co-exist and exchange.

Cheng closes his essay "cautiously optimistic, if not hopeful." That's exactly where I land too. The tools are extraordinary. The questions they raise are hard. The future is, as he says, hard to tell.

But the work itself — making buildings, serving clients, shaping places — hasn't changed in its fundamental orientation. What changes is the responsibility that comes with more powerful instruments. Use them well.

#BIM #Architecture #AI in AEC #DigiComm25 #Digital Twin #Revit #Sustainability #AEC Technology #Architectural Practice

Source article: Terry Cheng, "Working with Technology," Architecture New Zealand, Issue 4, July 2025 — published via Architecture Now. This post represents the author's personal commentary and perspective on the themes raised in that article.