Mike Rosen Architects: 17 One-Sheets Before the Event Ended

Mike Rosen Architects, an architecture practice, published a post on October 27, 2026 about putting an AI agent to work on real projects. Rosen writes that he met one-on-one with seventeen building-product suppliers at BOND Custom and that "before the event ended, every one of them had received a finished product one-sheet." An AI agent is software that takes a goal and carries out several steps toward it, not only one reply. A one-sheet is a single page that sums up a product or offer.

The setup Rosen describes is a good one. Turning a meeting into a finished page while the other person is still in the room is the kind of work that eats designer hours at a small firm. It is his own account, though. Rosen also sells AI services to architecture, engineering and construction firms, a group the industry calls AEC. For an owner, the better question is where a firm like yours should start.

That question matters because turnaround is something a client or supplier feels. A one-sheet that arrives the same afternoon reads differently from one that arrives two weeks later, after everyone has moved on. For a small firm, the gap between a meeting and a finished deliverable is often where follow-through quietly dies. Rosen's post is a picture of that gap closing. It does not tell you how wide your own gap is today.

Inside the agent: a production assistant with a brief

At a design firm, an AI agent is less a robot designer and more a production assistant that can follow a brief. In Rosen's description, the agent produced the one-sheets live during the meetings. For that to work, someone likely set up the format, the inputs and the review before the event started. The meeting supplied the details, and the agent assembled the page.

A deliverable here means any finished piece a client or partner receives: a one-sheet, a rendering, a proposal, a meeting summary. A rendering is an image of a design that shows how a space or building will look before it is built. Each of these has its own inputs, its own reviewer and its own standard for done. An agent helps most when those three things are already written down, even roughly.

That is the part that is easy to skip. Seventeen finished pages in one event points to a repeatable format, not a prompt typed in a hurry. If your firm has never written down what a good one-sheet contains, the agent has nothing steady to follow, and every page becomes a fresh negotiation.

Rendering hours at a large hospitality design firm

The second case comes from a much larger firm. Woolpert, a Google Cloud partner and therefore a vendor in this story, published a case study on WATG, a global hospitality design firm. Woolpert reports that WATG renderings "that previously took two to three weeks are completed in just hours." The page is undated, so treat it as a supporting benchmark rather than this month's news.

Woolpert also reports "the initial 50-user pilot yielding $755,000 in annualized savings." A pilot is a limited trial with a small group of users before a tool is rolled out across a firm. The same page says Gemini Enterprise is "producing the work of 5.2 full-time equivalents." A full-time equivalent, or FTE, is a way of counting work in units of one person working full time. These are Woolpert's figures, reported about a client, and they are self-reported.

WATG is a large global firm with fifty people in its pilot alone. That is not a promise about what a five-person studio will see. What the case does show is where the time was going. Rendering was a weeks-long step, and the firm put a clear number on it before and after. A small firm can borrow that habit without borrowing the scale.

Agency staffing at a European hotel group

A second vendor case comes from outside design. Twistag, a software vendor, published a case study on an unnamed European hospitality group on April 29, 2026. Twistag reports the group "was spending more than €1 million a year on agency staffing for shortfalls that internal capacity could have covered." Agency staffing means paying an outside firm for temporary workers to fill gaps in a schedule.

Twistag reports the group reached a "70% reduction in agency-temp spend." The client is not named, and the figures come from the vendor. It is a different business from a design firm, and it is not a small-firm case either. It belongs here for one reason. Twistag's account starts from one cost line, agency staffing, where work was going to outsiders.

For an architecture or design owner, the matching line might be outside renderers, freelance marketing help or late nights before a submission. You do not need a vendor's case study to find it. You need last quarter's invoices and an honest look at which deliverables kept slipping.

What the Bluebeam survey says about getting past a pilot

The broadest data in this story comes from Bluebeam, a construction software vendor owned by the Nemetschek Group. Its October 13, 2026 press release describes a survey of 750 U.S. AEC technology decision-makers, run online by Hanover Research in July 2026. It is a vendor survey, and the savings are self-reported estimates.

Bluebeam reports that "40% of firms running AI in production report savings exceeding $500,000." Production here means AI used in everyday project work, not in a trial. The same release says "only 28% have moved AI into production at all, and just 7% have embedded it across multiple core workflows." A core workflow is a process the firm runs on every project, such as drawings or design reviews.

Bluebeam also reports that "50% cite high implementation costs as the biggest barrier to scaling AI." Read together, the release describes an industry where plenty of firms are trying AI and far fewer have made it part of how projects run. That is the line between piloting and being ready to scale, and it is the line most worth knowing about your own firm.

Pick one deliverable and time it

Taken together, these sources suggest a first step. Before choosing a tool, pick one deliverable your firm produces again and again, and time it from request to finished. One-sheets, meeting summaries, early renderings and proposal sections are all good candidates.

Write down three things for that deliverable. What goes in, who reviews it, and what done looks like. If you cannot answer those in a few lines, that is the work to do first, with or without AI. If you can, you have the brief an agent needs, and you can test it on the next five requests instead of the next fifty.

Keep the test small and visible. One deliverable, one reviewer, a clear start and end time. That is how a pilot becomes something you can judge, rather than a subscription that quietly renews. It also keeps the decision in the owner's hands, where it belongs at a small firm.

Where Startup Miracle fits

Rosen's agent, Gemini Enterprise at WATG and Twistag's workforce software each did real work, by their own accounts. What a small firm usually lacks is a clear choice of which deliverable to hand over first, and a format written down well enough to hand it over.

That is the work Startup Miracle does with owners. We start with the jobs your firm already repeats, then help you decide which one is ready for an AI agent and which one needs a clearer process first. If you are weighing where to begin, our piece on how knowing where to start shapes the return on an AI tool walks through the same choice for a different product.

The same applies to quotes and estimates. We wrote about a dealer system that drafts a quote from photos, and the lesson carries over. Name who may send the finished page before the agent ever writes one.

Questions architecture and design owners ask

What is an AI agent for an architecture firm?

An AI agent is software that takes a goal and carries out several steps toward it, such as turning meeting notes into a finished one-sheet. It works best when the firm has already written down the format, the inputs and who reviews the result.

How do I start using AI at a small design firm?

Pick one deliverable your firm repeats, time it from request to finished, and write down its inputs, reviewer and standard for done. Then test an AI tool on the next few requests and judge it against that baseline.

How much can AI save an AEC firm?

Bluebeam's vendor survey reports that 40% of firms running AI in production report savings exceeding $500,000, as self-reported estimates. The same release says only 28% of firms have moved AI into production at all, so the first question is whether your firm is piloting or ready to scale.

If you want the long version of how we think about readiness, Start with a FREE quiz and Get Your AI Score.