There's a certain irony in using the metaphor of tracing paper—a technology architects haven't seriously used in decades—to sell them on the future. But when Illoca launched its Tracing Paper platform on May 6, 2026, the San Ramon, California startup wasn't really betting on nostalgia. The name is more conceptual: what if AI could act like that transparent overlay architects once laid over sketches, except instead of just copying lines, it generated production-ready floor plans and 3D models?
It's a compelling pitch, bolstered by $13 million in seed funding led by Bessemer Venture Partners. And it arrives at a moment when the architecture industry is simultaneously curious about and wary of AI's promises. The question isn't whether tools like this could save time—early experiments suggest they can. The question is whether they'll actually get used in the messy reality of professional practice.
Illoca's founders have a number they want you to remember: $720 billion. That's what they're calling the "Production Tax"—the annual cost of architects spending, according to the company's estimates, roughly 60% of their workweek (about 1,300 hours per person per year) translating design ideas into the technical files clients and contractors actually need. It's tedious work, the kind that makes people wonder why they went to architecture school in the first place.
Tracing Paper promises to collapse that translation layer. Sketch something rough, type what you want, mark up an existing plan—and watch AI agents build structured, editable models in real time. At least, that's the idea.
More Than Just Pretty Pictures
What sets Tracing Paper apart from the parade of image-generation tools flooding the market isn't the quality of its renders. It's what comes out the other end.
The platform accepts rough sketches, bubble diagrams, reference images, written descriptions, natural-language prompts, even redline markups. From those inputs, it generates editable 2D floor plans and 3D massing models that export directly into Autodesk Revit and Rhino—the software architects already use, for better or worse.
Illoca describes five core capabilities, though the language feels a bit like feature-list marketing. Augmented Sketch transforms rough drawings into editable floor plans. Adaptive Massing converts sketches or images into BIM-compatible 3D forms. Prompted Plans generates layout options based on text instructions. Agentic Refinement executes redlines and markups as actual design changes. Instant Facades explores facade variations without manual modeling.
The workflow splits into two modes: "Ideate" for creating from scratch, "Iterate" for refining. The company insists users don't need to be "prompt engineers"—it's built for professionals, not hobbyists trying to coax the right output from a chatbot.
Perhaps the most telling detail is the Revit integration. Illoca isn't trying to replace the existing CAD and BIM ecosystem. It's trying to work within it, which might be the only realistic approach. Credit costs for a Revit export range from 20 to 240 credits depending on complexity, according to the company's FAQ. That positions Tracing Paper as a preprocessing layer—not a wholesale replacement for tools architects already know.
The founders frame it as returning "creative agency" to architects. Less time wrestling with software, more time making design decisions. The company claims the platform could turn "weeks of design work into days," though that's the kind of claim that requires real-world validation, not just company testimonials.
The Economics of Credits
Tracing Paper launched with a freemium model that might feel familiar to anyone who's used AI tools lately.
The free tier includes 100 credits that refresh monthly—enough to experiment, probably not enough to run a real project. The Explore plan costs $29.99 per person per month and bumps that to 3,000 credits, plus early access to new features. A Team plan at $59.99 and up per team per month adds pooled usage and collaboration tools, though it's currently waitlist-only. Enterprise pricing is custom and includes the usual enterprise features: custom agents, security controls, the works.
New sign-ups get 1,000 starter credits to play with. Costs vary: image generation runs 10 credits, agent chat costs 10 credits per turn, agent edits range from 5 to 50 credits depending on scope, floor plan generation runs 15 to 150 credits, and 3D massing costs 10 to 120 credits.
The company says users own their designs outright. Illoca's privacy policy, effective May 5, 2026, promises the company won't use proprietary user inputs—sketches, prompts, project models—to train its models. Enterprise customers can opt for zero-data retention and custom-trained models, which matters in an industry where client confidentiality is non-negotiable.
Whether the credit-based pricing aligns with how architecture firms actually budget and staff projects remains to be seen. Studios don't think in credits. They think in billable hours, project budgets, and overhead.
A Crowded Field, Measured Enthusiasm
Illoca isn't entering virgin territory here.
Autodesk Forma already handles early-stage generative planning within the Autodesk ecosystem. TestFit offers AI-assisted site planning and feasibility studies. Maket AI rebuilt its platform in 2025 with agentic features for floor plan generation. Finch3D provides generative tools for floor plans and massing. Hypar offers a cloud platform for publishing generative building logic.
The broader architecture industry has shown interest, but it's measured. A March 2025 study from the American Institute of Architects found excitement tempered by real concerns about adoption—liability, accuracy, the learning curve. A December 2025 Bluebeam survey suggested the AEC sector remains slower to adopt AI than, say, tech or finance, though early adopters reported significant time savings. Nearly half reclaimed 500 to 1,000 hours on certain tasks, according to that survey.
Illoca named Kajima Corporation, the Japanese construction giant, as an early partner. Chief Architect Yasuhiro Nakano offered a quote in the press release about faster iteration while preserving design intent. But the company hasn't yet published detailed case studies or independent user reviews, which would offer more insight into how the platform performs under real deadlines.
The Founders' Bet
Chin-Yi Cheng and Chiaowei Yu aren't exactly outsiders to this problem.
Cheng previously worked at Google DeepMind and Autodesk's AI Lab—exactly the kind of background you'd want for building AI tools aimed at design professionals. Yu led BIM efforts at Tesla and worked on complex healthcare projects, which means he's familiar with the gap between what AI demos promise and what production workflows actually require.
In the announcement, they described Tracing Paper as "the interface they spent their careers wishing for." It's a personal framing, the kind founders often use when they're selling a product born from frustration.
The $13 million seed round included AIX Ventures, Root Ventures, and Alt Ventures alongside Bessemer. Bessemer's May 6 blog post sketched a longer-term vision: professionals generating "code-compliant, construction-ready designs in the amount of time it takes to open a file."
That's aspirational, to put it mildly. Tracing Paper as launched focuses on early-stage schematic design—not the fully engineered construction documentation that requires coordination with structural engineers, MEP consultants, and code compliance checks. Getting from concept to construction-ready plans involves layers of technical complexity that can't be hand-waved away with AI, at least not yet.
The Test Ahead
As of early May 2026, Tracing Paper is live. Free sign-ups are available, paid tiers are active, and the Team plan is accepting waitlist applications.
Coverage so far has been largely company-controlled—press releases, the official site, investor blog posts. Third-party reviews and hands-on evaluations haven't surfaced publicly yet, which means much of what we know comes from Illoca's own framing.
The real test will be adoption. Not whether architects can use the tool, but whether they will—whether the outputs are reliable enough to integrate into existing workflows, whether the credit pricing makes sense against studio economics, whether the time savings materialize in practice and not just in demos.
Early architectural AI tools have stumbled over the gap between impressive demonstrations and production-ready reliability. Models that look great in screenshots fall apart when you need them to coordinate with mechanical systems or comply with local building codes. Illoca is betting that tighter integration with Revit and a focus on structured, editable outputs—rather than flashy renderings—will close that gap.
It's a reasonable bet, though not a guaranteed one.
For now, Tracing Paper offers something like a proposition: maybe AI can handle the tedious translation work that currently eats up most of an architect's week, freeing up time for the design decisions that drew them to the profession in the first place. Whether that happens in practice, or whether it's just another layer of software architects have to learn and bill for, will become clear soon enough.
