AI production intelligence for architecture, engineering, and construction
Claude Code–level production intelligence for architecture, engineering, and construction.
Run prebuilt work packs that draft common deliverables and route them to the right reviewer.
What we're building
Project Cornerstone is building a private, permissioned AI production layer for architecture, engineering, and construction.
Built on the tools the industry already uses: Revit, ACC, and PDFs stay the source of truth.
One permissioned context built from your past projects, scoped to each party, from developer to trades.
Run prebuilt work packs that generate first-pass deliverables, routed to the right reviewer, grounded in your project sources.
The whole value chain
Each party works from the same permissioned context.
Developers. Fewer change orders and schedule surprises.
Architects. More projects with the same team. Better QA.
Engineers. Assumptions verified early. Cleaner first passes for review.
GCs. Design intent on demand. Faster RFI cycles.
Subs and trades. Scoped field answers and takeoffs, tied to the right sources.
Designed so the value grows as more of the chain joins.
On top of your tools
It sits on the tools the industry already uses. Revit, ACC, and PDFs stay the source of truth.
Generic AI tools don't have your project context. Project Cornerstone answers from it, with citations back to source.
- Retrieval: start from what the firm already built
- Quality: catch issues before a set goes out
- Constraints: surface code and standards while change is cheap
- Checks: confirm consultant assumptions before they harden
- Handoff: downstream teams see only what they need
Self-serve by design: start with a few work packs and expand from there.
The goal is not to replace architects, engineers, contractors, or tradespeople. The goal is to give them a trusted AI layer that preserves context, accelerates repeated work, and supports human-reviewed decisions across complex projects.
Demand swings. Payroll doesn't have to.
In the AEC industry, work arrives in waves, and teams can't staff up and down fast enough in response.
Every new project rebuilds context someone in the chain already has.
With elastic inference, teams handle peak workload without permanently staffing for the peak.
Some labor spend shifts to usage-based AI spend, closer to a variable cost.
"We would not have to staff up and cut back every time the market turns. For a firm our size, that is the game changer."
CFO, US multifamily architecture firm
Workflow packs
Up to 30% of a phase's spend is work that repeats from past projects. Packs turn it into repeatable, reviewable workflows.
- CD-to-DD pull-forward. Prior detail work drafted into the current set.$52.5K to $125K modeled savings per project
- Engineering draft outputs. First-pass deliverables routed to engineer review.$75K to $160K modeled savings per project
- GC and trade estimates. Quantities and first passes for estimator review.$87.5K to $250K modeled savings per won project
- Drawing QA. Small errors caught before the set goes out.being measured in production trials
Modeled on a benchmark composite project · assumptions available on request
Outputs route draft → reviewer → approved deliverable, with permissions and audit.
Define. Configured around your standards and review gates.
Run. On live project work, as often as needed.
Review. Every output reaches the right reviewer first.
Packs are designed to run self-serve, with expert support when you want it.
Your knowledge stays yours
Each party's data and outputs stay that party's property. Nothing trains public models without permission.
Access is scoped by source, role, and project. Sharing across the chain is permissioned down to each source.
Answers include citations back to the underlying drawings and documents.
| Source | Architect | Engineer | GC | Trade |
|---|---|---|---|---|
| Firm standards | Full within role scope | Not shared | Not shared | Not shared |
| Model and drawings | Full within role scope | Scoped view | Scoped view | Scoped view |
| Consultant inputs | Scoped view | Full within role scope | Scoped view | Not shared |
| Review history | Full within role scope | Scoped view | Not shared | Not shared |
| Construction evidence | Scoped view | Scoped view | Full within role scope | Scoped view |
Where the work stands
Now. Completing a paid assessment with a design-partner architecture firm.
Next. A pilot on a live project, architecture side first.
Goal. The full value chain on the project, with developer buy-in.
Work with us
We want to hear from:
Firms, developers, GCs, and trades with repeated production work
Investors in vertical AI and AEC automation
Also hiring (builders): AI + full-stack engineers, computational designers, BIM experts.
An email gets you a working session on one of your live projects.