BIM MONKEY
A year of case studies from live Revit models at our firm: sheet setup, template cleanup, code checks, and the parts that still need a person.
Search for “AI for Revit” and you’ll find vendors promising CD sets in minutes, models that design and then draw themselves, and DIY tutorials on setting up your own MCP server in a few simple steps.
This post is different. It’s a series of case studies from the past year of using an AI assistant inside live Revit models at our firm: what worked, and where a person still has to step in. Every example is from a real project.
We use AI inside Revit as a BIM assistant. It reads and writes to the model, handles the repetitive day-to-day tasks, and checks before it changes anything. We’ve also used it to plan and build out our BIM standards, finally getting to the items that sat on the “it would be nice to eventually…” list for years. It still needs you for the drawing decisions.
The most common question we get from other architects is some version of “can it set up my sheets?” Yes, with limits.
What works: populating a cover sheet from project data. On a recent residential job, the assistant filled in scope, address, parcel number, zoning designation and the environmental review note from the project record, read the surveyor’s name and license number off a CAD title block image, and caught that the template still cited the commercial code on a residential project. The user knew the process and pointed the assistant at the slow part: gathering the source documents in one place and copying them into the formatted blocks in the Revit model. The site plan submission needed to get a permit number was ready in about thirty minutes. On a good day that used to take two hours. At thirty minutes, it fits between other tasks instead of waiting for an open afternoon, and the first step of the permit process doesn’t slip a day or two.
What also works: placing viewports on sheets, aligning them to the title block, and fixing the view labels afterward, because the first placement is usually off by the width of a label.
What does not work: deciding what belongs on a sheet. That’s still a drawing decision. The assistant will place the viewports you name.
Years of CAD imports leave a template full of line styles nobody dares delete. One of ours had 107. Deleting by hand is slow because every style has to be checked for use first.
The assistant checked every style for references before touching anything, deleted only the 42 with zero uses, and built out the missing text size ladder (a set of text sizes that scale together) while it was in there. The result was 65 line styles, all firm standard, nothing broken. It then wrote a short list of what the template still lacked, in priority order.
The order of operations is the point. It checks, faster and more thoroughly than you or I could, reports back with enough information to make a decision, deletes on your approval, then checks again to confirm the result. A tool that deletes first isn’t an assistant. It’s a liability.
New file, missing the firm’s filled regions, dimension types and text styles. With the sibling project open in the same session, the assistant copied 38 filled region types and two dimension types across, then renamed 26 legacy prefixed types to the current standard. Under a minute, zero failures.
This is the kind of task that makes the case for AI in Revit better than anything flashy. It’s tedious, error prone by hand, and it happens on every project. That includes the old ones you no longer enjoy opening. Do a Save As, update the standards, and you’ll be glad you did.
The plan reviewer wrote “demonstrate compliance with the solid waste section in the plans.” The zoning code has been revised recently to allow additional units on a site, and the user wanted a refresher on what it currently requires. The assistant pulled the current municipal code from the city’s own site and laid out the options the architect could choose from to meet it. Once the user drew the revisions, the assistant checked the enclosure against the requirement and wrote site plan notes with container counts and sizes for review. It also flagged that the enclosure sat where the code doesn’t allow it, and drafted a note requesting the alternative arrangement the code provides for, instead of pretending the plan complied.
That last move is the difference between a tool and a colleague. It put the real conflict in front of the reviewer on purpose.
Most firms our size will never have a dedicated BIM manager or a Dynamo or pyRevit expert. An AI assistant fills that role for everyone with access to it. With a little coaching, it reads and writes to the Revit API directly, so the limit is what the API allows, not what you can script.
At first it can feel daunting. What do you ask it? What’s the best use of your time, and its time? Stick with it and the ideas start to flow. Within a few sessions you gain confidence and BIM Monkey learns how your firm works. Soon it’s clouding revisions, auditing models, building families, organizing and renaming, and building firm standards alongside you.
The assistant is fast, thorough and doesn’t get bored. It doesn’t know what the project needs. It will place every viewport you name, but it won’t notice the wall section you forgot to cut. It will lay out three ways to meet the solid waste code, but picking the one that works for this site, this client and this budget is still your call. It will clean a template down to firm standard, but someone had to decide what the standard should be.
That’s not a gap waiting for the next software update. It’s the job. Any tool that claims to make those decisions for you is either overselling or making them badly. The best use of AI in Revit is clearing out the hours of checking, copying, renaming and filling in, so the time you spend in the model goes to the decisions only you can make.
Barrett Eastwood is Co-Founder and an Architect with Wood Studio in Seattle.
Eric Chrabot is Co-Founder of BIM Monkey Incorporated.