Motif doesn't have an automatic dimensioning tool in the sense a BIM 1.0 system does. Instead, dimensioning is fully configurable through the agent. You tell it how your firm dimensions, and it dimensions that way. I mentioned this in my post from Rhino World.
That's the right call, because every practice dimensions differently. Grid to grid or grid to face. To structure or to finish. Running dimensions or chains. Which strings belong on the general arrangement plans and which are left to the details. An automatic dimensioning tool has an answer to each of those, and they're someone else's answers. Most of us have spent the afternoon after an auto-dimension command moving, deleting and redrawing strings until the sheet looked like one of ours.
Motif handles this with skills, an idea it has taken from Anthropic's Claude. A skill is an instruction written in ordinary language rather than code. In Motif, skills can belong to a person, a project or the company, and one that a single person develops can be promoted into a company standard. Amar Hanspal, Motif's CEO, compares them to AutoLISP.
AEC Magazine's look inside Motif in September named the problem that comes with them. Two skills may do the same task differently for different building types, offices or clients. It called this "the old CAD standards problem applied to agent behaviour", and noted that Motif doesn't have a complete answer yet.
It's the old CAD standards problem because it is your CAD standards. The agent can only dimension the way your firm dimensions if it's been told how.
The same goes for content. People have complained for years that Revit and Archicad are closed and inflexible. A new product means waiting for someone who can build a family, write a GDL script or put together a Dynamo routine. At Rhino World I was talking to Motif about how it can take any manufacturer's data sheet, a new light fitting for example, and have AI create the object with all the parameters it needs.
I said that could get out of control in an office pretty quickly. Matt Jezyk, Motif's VP of Industry Solutions, agreed. Some practices, he said, had flipped their view on openness once they saw the potential for chaos.
The experts were never just a bottleneck. They were where the standard got applied. Every family went past someone who knew how the office named its parameters, which ones it scheduled and what a light fitting had to carry. Take them out of the loop and that knowledge has to live somewhere the agent can read it. Otherwise you get a dozen versions of the same downlight, each with its own parameter names.
Either way, your standards have to be written down.
Most firms have them. A CAD or BIM standards manual, a drawing standards document, and a template that holds more of the rules than either. Fortunately LLMs are good at reading them.
They don't read them the way people do, though. People skim the manual once and learn the rest from whoever sits next to them. Where it's vague, out of date or contradicts itself, they use judgement and nobody notices. An agent reads all of it, every time, and does what it says. The gaps people have been quietly filling for years show up on the first sheet.
If your standards aren't written down, or the document hasn't kept up with the template, AI can help with that too. Point it at a set of issued drawings, the template and a folder of typical details, and ask it to describe what the firm actually does. How dimensions are set out, how text and tags are styled, what goes where on a sheet. What comes back is a first draft of a standard that describes practice rather than intention. Someone senior still has to read it and decide which inconsistencies are deliberate.
The same habit turned up at Rhino World. McNeel's developers run their overnight crash process from instructions written in a markdown file. After using RhinoAI to build a furniture schedule and push the data back into the model, Scott Davidson said on McNeel's forum that he'd like to write a RevitAgent.md, a set of instructions for the agent.
Once the rules are written down, the agent shouldn't have to reinterpret them on every sheet. Brian Mathews, Motif's CTO, talks about "not renting the same thought twice". Use the model once to produce a skill or some code, review it, then run that. Davidson's line at Rhino World, "AI is probably right", is fine for a first pass at room names. It isn't fine for dimensions on an issued drawing.
Written standards are also what it now takes to try a new tool. Motif, Qonic and the agents arriving inside Rhino, Revit and Archicad through MCP can only be judged properly on your drawings, to your standards.
Evaluating a new BIM tool used to mean rebuilding the template, the families and the title blocks before you could get a fair answer. That took weeks of your best technical people, and it's a large part of why so few firms ever ran a serious comparison. Standards written in plain language travel. The same document can go to each tool, and how well each one follows it is the test.
So the work to do before choosing anything is extracting your standards, organising them and cleaning them up. It pays off whichever tool you end up with, including the one you already have.
Start with dimensions. Every firm thinks it knows how it dimensions. Writing it down is how you find out.