Where building plans meet retail decisions
Architects work with geometry and dimensions. Space planners allocate shelf space to product categories. Merchandising and store operations teams decide how that space should be used. Each discipline brings a different perspective to the same store. AIsly gives them a shared floor plan to work from.
We wanted to make existing CAD drawings usable for collaborative planning in a browser. Each fixture becomes an object with dimensions, a category and a position. Teams can discuss changes, analyse space and compare layouts around those objects. We developed AIsly independently and refined the prototype using real plans and feedback from planning specialists.
Our engine recognises shelving in the CAD geometry
Much of the development sits behind the import process. Our engine reads a DXF file, identifies shelving structures and proposes editable objects, preserving their position, orientation and real dimensions. A person reviews and confirms the result.
The retail team can then move, rotate, add to and categorise those fixtures. Each object connects a physical position in the drawing to the information needed for space planning. PDF plans can also serve as a reference layer, while automatic fixture recognition works from the DXF geometry.
A layout decision keeps its context
Suppose a product category needs more space. A planner adds a row of shelving, assigns its category and checks its position and the surrounding aisle widths. The same working plan provides the shelf-length breakdown by category. The physical layout and the merchandising decisions stay connected.
Questions can be attached to the relevant fixture or a precise location on the plan. Replies and their status remain there too. Other participants see changes as they happen, and named versions preserve alternative layouts. A frozen version records an agreed decision without being overwritten by the next round of edits.
Walk through the layout you have just planned
What will a row of shelving look like from the entrance? What can someone see from one part of the store into another? AIsly renders the layout from the 2D editor as a 3D space. There is no need to rebuild it in a separate modelling application.
The same plan is available in VR. Users choose a store plan and open it directly in the headset’s browser. We developed and tested this WebXR functionality on the Meta Quest 3. A team can experience the proposed space while changes are still being made on the plan.
The agreed layout goes back into CAD
Once the team has agreed on a layout, AIsly can export it as a DXF layer at its true dimensions. Architects and building specialists can continue working with it in their existing tools. Detailed shelf planning and planograms remain part of the subsequent workflow.
AIsly also retains a structured record of stores, floor areas and fixtures. Sales floor area and shelf length by category can be analysed. Versions, comments and responsibilities stay associated with the relevant store, so a decision can be understood later in the context of the layout it concerned.
Ready for a focused pilot
AIsly is a working beta. Its technical foundation combines a React and TypeScript interface, three.js for 3D, and a FastAPI backend with PostgreSQL. The next step is a limited pilot involving a real planning exercise, to assess how the interface, team coordination and CAD handover fit into day-to-day work.
The structured data creates a basis for future comparisons between stores, automated checks and AI-assisted planning suggestions. A tablet AR view is also in preparation. Those are future extensions. Import, collaborative planning and VR walkthroughs are already implemented.
AIsly keeps the building geometry and the retail decisions together. Each fixture retains its dimensions, category and position as the team works on the layout. That same plan supports space analysis, a VR walkthrough and a DXF handover.