TT5: Backlog Busting
October 15, 2026
This weekly email shares my curated selection of interesting, useful topics related to BIM, design, and technology. I aim to share useful information and a few laughs along the way. So, what did I find interesting this week? Read down to find out!
A common thread across this week’s links is AI that does the work instead of just describing it. Agents are going into Revit and clearing out the backlog: 48 state maps in an afternoon, 201 panel schedules in half an hour. Before we get to this week's articles, here are two quick ones from me. First, a short video of my recent AU presentation on how we're using AI at Ulteig. And second, I recently rebuilt my sketch-to-Revit workflow based on what I've learned since the first version I created earlier this year. The tools are getting really good, really fast. But for best results, you still need to understand your process.
Alright, here are five things to check out this week.
#1: Cleaning up the Waybacklog With Claude
Every BIM manager has that task that's been sitting on the to-do list for years. Jason Kunkel's was a consistent set of state vicinity maps. Using Claude and his own Revit MCP server, he built all 48 lower states as families in about four hours, complete with county linework, toggleable county fills, interstate labels, and capital markers. The best part is how he got there: he ran the first state manually, iterated with Claude on Alabama until it was right, and then let it rip.
Click here to read the full article at Design Tech Field Guide.
#2: Using AI With Panel Schedules in Revit
Design Master connected an AI agent to Revit through their MCP server and had it balance loads, reorganize circuits, and clean up descriptions across 201 panel schedules in 32 minutes. Along the way, it flagged a panel feeding devices in the wrong dwelling unit. The real lesson is in the workflow. Perfect the process on one panel, have AI write its own instructions, generalize them, and reuse them on a different project. Through it all, the engineer still makes the judgment calls.
Click here to read the full article at Design Master.
#3: The Cost of Every Compromise
Portugal's largest architecture-only practice spent two years testing AI rendering tools, and most failed for one reason: geometry drift. They put in a concept, and something "completely wild" came out. They settled on Visoid because it kept the design, and now they use it to show clients what a material swap or faƧade change actually does to a building before they lock in the decision. The 80-90% time savings are nice, but the real value is in catching cost-driven compromises while there's still time to push back.
Click here to read the full article at Architizer.
#4: Beyond Proposal Automation
Nicole Chavas of Greenprint Partners, a 35-person firm with no dedicated knowledge management team, is redesigning work around AI instead of bolting it onto existing processes. She walks through two efforts: a pursuit process that draws on institutional knowledge instead of reinventing it for every proposal, and a training pipeline that turns interviews with subject matter experts into reusable learning content. If you're at a smaller firm where all the expertise lives in a few people's heads, this one's for you.
Click here to read the full article at Knowledge Architecture.
#5: Reconnect Academia and Architecture Practice
Ramiro Diaz Granados of SCI-Arc argues that schools and offices have drifted so far apart they now operate as separate professions. His fix is a shared "disciplinary core" of six questions that only architecture can answer, covering tectonics, experience, type, the city, representation, and nature. It's a dense read, but his critique of renderings made to sell rather than to think with is a useful counterweight to the AI-everything theme this week.
Click here to read the full article at The Architect's Newspaper.
That’s all from me. I hope you’re having a great week.
Michael