Allsite.ai ✓

Allsite.ai is an AI-powered civil engineering design platform built around two design agents -- Level AI for grading design (roads, lots, retaining walls) and Service AI for pipe network design (storm and sanitary systems) -- that integrate directly into Autodesk Civil 3D, producing the same native Civil 3D objects an engineer would create by hand rather than a separate, disconnected output format.

The company was founded by Sam Blackbourn (a Chartered Professional Civil Engineer who also owns and runs a large civil land-development consultancy) and George Stirling (a software engineer with a background building secure, scalable systems), with an advisory board that includes Bronson Johnson (former Head of Infrastructure and Sustainability on the California Forever Project and former Senior Director at Sidewalk Labs) and Don Stock (former CEO of Naylor Love) -- a US- and New Zealand-based team per the company's own about page.

According to Allsite.ai, the underlying technology has been in development for roughly seven years, and Service AI designs have already been constructed in the New Zealand market, including a site that withstood a 200-year flood event without damage; the company states its platform has been used across more than 15,000 dwellings to date.

The core workflow is genuinely automated at scale: users upload DXF files (road centerlines, curbs, lot boundaries, retaining walls, existing drainage and sanitary systems), and the AI runs on cloud GPU infrastructure to return an optimized, code-compliant grading and/or pipe-network design -- a single optimization can take around 2 hours for a typical site and up to 48 hours for mega-developments exceeding 10,000 lots, per the company's own FAQ.

Published case studies name Stantec, Civix, Brookfield Properties/Brookfield Developments, Inertia Engineering and Wood Rodgers among users, with results cited including cutting preliminary design hours by roughly 25% on a 60+ lot infill site, balancing a 400-acre, 1,500-lot California community's earthworks, and identifying an earthworks solution that reportedly saved approximately $2.5 million versus an initial design surface on a Queensland project -- Allsite.ai's Inertia Engineering collaboration also won Best New PropTech 2026 at the Urban Developer Awards for Industry Excellence.

Pricing is entirely usage-based rather than a flat subscription: an Ad Hoc pay-as-you-go option requires no bundle, a Pro bundle runs $10,000 with a 10% token-rate discount, and an Enterprise bundle runs $100,000 with a 20% discount -- the company's own pricing calculator gives illustrative per-run costs (for example, roughly $405 per Level AI run and $41 per Service AI run on a 200-acre data-center example under the Pro plan) that scale with project size and complexity.

Editor’s takeBIM•STACK editor's take: Allsite.ai's real differentiator is running AI grading and pipe-network optimization on genuinely large compute (the company describes single optimizations running for hours on clusters of NVIDIA GPUs) while still outputting native Civil 3D objects that plug directly into an engineer's existing quality-assurance process, rather than a black-box result disconnected from familiar tools. The usage-based, per-run pricing model means cost scales with project size and complexity -- firms should use the vendor's own calculator against their specific project size before assuming Ad Hoc, Pro or Enterprise is the right fit.

Because Level AI and Service AI output native Civil 3D objects rather than a separate proprietary format, engineers already fluent in Civil 3D face a relatively short learning curve for interpreting and refining the results -- the review and quality-assurance process largely mirrors how they'd already check a manually-produced grading or pipe-network design.

The upload workflow itself (submitting DXF files for road centerlines, curbs, lot boundaries, retaining walls, and existing drainage/sanitary systems) requires attention to input-data quality, since the company's own FAQ notes that inputs are validated with feedback generally provided within 24 hours -- clean, complete source DXF files produce a smoother first run than incomplete or ambiguous inputs.

Teams should expect an initial calibration period -- similar to the side-by-side validation approach case studies describe (running the AI output against manual CAD grading to build confidence) -- before relying on the AI-generated design as the primary basis for construction documents rather than as a fast first pass for engineering judgment to refine.