AI for constructionShipped

Lift

Complete structural steel takeoffs in minutes, not hours.

Role
Design, UX, UI and interactive prototype.
Team
Client engineering built the detection model and the app
Stack
Figma design and interactive prototype; cloud web app
95–99%
detection accuracy on beams, columns, grid, bolts, connections and trusses
Product accuracy claim, as published by SketchDeck.AI
Minutes
per takeoff where the manual count took hours
Product claim

The problem

Structural steel estimating teams count beams, columns, bolts and connections by hand from multi-page PDF drawings. It is slow, repetitive and error-prone, and the estimate is only as good as the count.

Lift uploads the drawings, analyzes and quantifies the members, and gives the estimator a measurements panel to review and export.

Key decisions

Constraint

An estimator will not sign off on a count they cannot check.

Options

Auto-commit the AI count, or make confirmation the core loop.

Decision

Detected elements are shown as accepted, missing or rejected labels for the estimator to confirm, page by page, before anything is counted.

Trade-off

Slower than fully automatic. Trusted, which is what gets it used.

Constraint

Real drawing sets are many pages at different scales.

Options

One page at a time, or a project-level workflow.

Decision

Projects hold drawings and floor plans; the user selects pages to analyze, sets scale and calibration once, applies it across pages, then iterates with add and edit.

Trade-off

More setup up front, less rework later.

The work

Lift takeoff screen
Takeoff reviewAI-detected elements with labels to confirm.
Lift screen
Lift screen
Lift screen
Lift screen
Lift screen
Lift screen

Where AI was used

The product is AI detection with human confirmation. Design work was done in Figma.