The Third Tooling Shift of My Career
What AI Actually Does on My Drafting Table Today
I’m seventy-four years old. I started drafting on a board in 1974, watched the industry move to CAD in the late 1980s, lived through the BIM transition in the 2000s, and I’m now in the middle of the third major tooling shift of my career.
The AI one.
I want to write this piece carefully, because there is a lot of bad writing about AI in architecture right now—and most of it is being done by people who have never stamped a drawing in their lives. So instead of another think-piece, here is an honest inventory: the AI tools that actually sit on my desk today, what they do, where they fail, and what I’ve concluded about the shift after building inside it for the last two years.
The Three Tools I Actually Use
- AECLogix Proof — a Revit plug-in for keynote checking. This is one I’m building. It does one specific thing: it checks the tagged keynotes in a Revit model against the actual project specifications and flags every mismatch. A keynote pointing to a spec section that doesn’t exist. A spec section that nothing in the drawings actually references. Every gap between what the drawings say and what the spec book says. Why does this exist? Because every architect reading this knows that keynote-to-spec coordination on a typical project is held together by hope, the project architect’s memory, and a Friday-afternoon spot check. The errors are real, they’re embarrassing in CA, and the cost of finding them by hand is too high to do it consistently. Proof finds them in seconds. The honest part: it doesn’t fix anything. It tells you what’s broken. The architect still has to decide what to do about it. That’s the right division of labor.
- COI Autopilot — subcontractor certificate-of-insurance verification. Also something I’m building, on the AECLogix side. This one is for general contractors more than for architects, but I include it here because it’s the clearest example I know of where AI does a job no human wants—and no software has previously done well. A mid-size GC tracks hundreds of subcontractors, each of whom needs to maintain current certificates of insurance with the right limits, the right additional insureds, the right endorsements. The current process is a part-time staffer with a spreadsheet, chasing PDFs, calling brokers, and missing things. COI Autopilot reads the COIs, extracts the data, compares it to the project requirements, and flags what’s wrong. The lesson from building it is this: AI is at its best when you point it at a job that already exists, that everyone hates, and that no one is doing well. It is at its worst when you point it at a job that doesn’t exist and try to invent a use for it.
- The Ghostwriter Desktop — my drafting partner for everything that isn’t a drawing. This is the one I use personally, every day. Client letters. Spec language. The first draft of this newsletter, which I then mark up the way I’d mark up a junior architect’s writing. Email responses to owners that I can’t bring myself to write at 7 p.m. on a Thursday. The mistake people make with this category of tool is treating it as a writer. It isn’t. It’s a draftsman. It produces a draft. The architect’s job is to redline it—the same way I redlined drawings for forty years. The output is only as good as the redline. I’d estimate it saves me four to six hours a week on writing tasks that used to eat my evenings.
What It Doesn’t Do
Now let me be clear about the boundaries—because this is the part the marketing people leave out.
AI does not design. It produces options, sometimes good ones, but it has no judgment about which option is right for this owner, this site, this code climate, this budget. The architect still designs.
AI does not understand the project. It understands the document in front of it. The architect still has to hold the whole project in their head.
AI does not replace the cheap part of practice—the drawing, the typing, the formatting. That part has already been getting cheaper for thirty years. What AI replaces is the medium-cost coordination work that used to sit awkwardly between the cheap tasks and the expensive judgment. Keynote coordination. COI tracking. First-draft writing. Schedule cross-checks. That entire layer is collapsing fast.
The expensive part—judgment, taste, the ability to look an owner in the eye and tell them what’s wrong with their idea—is not collapsing. If anything, it is getting more valuable. Because every architect now has a coordination layer that runs ten times faster than it used to, the constraint on the practice moves up to judgment.
---
Below the paywall: my prediction for which architects win the next ten years, the bet I’m making with AECLogix, and The AI Tools I Actually Use — the one-page Redline that lists every tool currently on my drafting table, the prompts I run them with, and the workflows that produced the four-to-six-hour weekly time savings I mentioned above. Steal the ones that fit your practice. Skip the ones that don’t.
If you’re on the free tier and this Field Note has earned its way into your week, the paid tier ($8/month, $80/year) is what makes the Redline series possible. Every Tuesday Field Note ships with a printable working tool like this one.
What I Think Comes Next
A prediction, with the caveat that I have been wrong about technology transitions before.
The architects who win the next ten years will not be the ones who learn the most AI tools. They will be the ones who understand the practice deeply enough to know which tasks should be delegated to AI—and which absolutely should not.
Put another way: AI in AEC is going to be won by people who understand the practice, not by people who understand the tools. The tools change every eighteen months. The practice doesn’t.
This is also, frankly, the bet I’m making with AECLogix. I’m not building AI products for the AI-curious. I’m building them for working architects and working contractors who already know exactly which parts of their day are wasted on coordination they shouldn’t be doing. The product is the time it gives them back.
The Honest Part, Again
I am not an AI evangelist. I am an architect who has watched three tooling transitions and has opinions about how this one is going to play out.
Here is what I would tell a thirty-year-old architect today: don’t ignore this, don’t worship it, and don’t let anyone tell you it’s a substitute for understanding the work. Pick up the tools that solve the problems you already have. Put down the ones that don’t.
Stay close to the practice.
That, I think, is what every previous tooling shift has rewarded. I don’t see any reason this one is different.
---
📄 Redline — The AI Tools I Actually Use
A one-page inventory of every AI tool currently on my drafting table, what each one does, the specific prompts I run them with, and the workflows that pulled four-to-six hours back into my week. Three tools you’ve probably heard of, three you haven’t, and a candid section on the tools I tried and abandoned.
Steal the ones that fit your practice. Skip the ones that don’t. Update it as the tools change — I will.
— Randy