Revit answers questions about the building. Point at a wall and it tells you the make-up, the area, which phase it belongs to, what it costs. That's been normal for fifteen years and nobody finds it remarkable.

Now ask the same kind of question about the practice. How many of the eleven live jobs are past the issue date on their current stage. Which three haven't had a client decision in a fortnight. Which consultant is currently holding up more than one project.

Same shape of question. Nothing answers it, and there's no reason it shouldn't.

That gap isn't about how any firm is run. It's about where the software industry pointed. Fifteen years of development went into how drawings get made and almost none of it reached the practice around them. Architecture was sold extraordinary tools for making the thing and almost nothing for running the firm that makes it, and the firms that noticed did what you did, which was assemble something workable out of the tools that existed.

AI has so far been pointed at the same place. Generation. Floor plans, massing, elevations, renders that used to take a night.

Which is the part a good firm is already strong at. You have a studio full of people whose judgment on a plan is better than anything software will produce, and the value of making that faster is real but it's bounded. It makes a capable team quicker at the thing they were already good at.

The question worth asking at your size is different. Eleven jobs, a team of thirty, commercial clients who expect an answer the same day. What would it be worth for the practice itself to be askable.

Not a report somebody prepares on Friday. An answer, now, the way Revit gives one.

There's a short list of questions a principal shouldn't have to ask a person. Where everything stands. What's waiting on somebody outside the firm. What's moved since Monday. Which job is quietly taking the most senior time this month. Any firm running double figures of live projects should be able to get those in seconds, and the reason it can't yet has nothing to do with the firm.

Here's the part that decides whether an assistant is worth anything.

You already run good tools. Something for drawings, something for messaging, something for documents, something for accounts. Each one answers well about its own slice, and each one is probably a sensible choice.

None of them answers across two.

Put an AI assistant on top of that arrangement and it inherits exactly the same boundary. It can tell you what the drawing system knows, then what the accounts system knows, and leave you to do what you were already doing, which is holding five partial answers in your head and working out what they mean together. The assistant is only ever as broad as the system underneath it.

Which makes the system the decision, not the assistant. A practice held in one place can be asked a question that crosses stages, money, people and time. A practice spread across five places can be asked five smaller questions, and the joining up stays with you.

There's a standard worth holding an assistant to, and it isn't caution. It's the same bar you'd set for a competent new hire. It answers from the practice's own records rather than from general knowledge. It acts only where you've said it can act. Anything it changes is visible, attributed and reversible. A good assistant makes the work easier to carry out without making responsibility harder to locate, and any tool that can't meet that isn't ready for a firm with your liabilities.

If you want to test this on your own practice, it takes ten minutes and no software. Write the five questions you'd most want answered without having to ask anybody. Be specific, use real projects, and include at least two that cross between money and progress. That list is a specification. It tells you plainly what your current arrangement can and can't do, and it's a sharper measure of a tool than any demonstration.

Projectsmate holds the practice in one system, projects, teams, time and the financial picture together, and Iris answers from it, so those questions get asked of the practice rather than assembled from four places.

The building has been answerable for fifteen years. The practice can be too, and the firms that get there first won't be the ones running the most tools.