Mach9’s AI CAD software now works on data captured from the air. The company’s news index carries the announcement, headlined “Mach9 now works in the sky” and dated 4 November, under the summary “Turn data from drones, helicopters, and planes into finished survey deliverables faster with Mach9’s AI CAD software.” It lands two weeks after the company raised a $34 million Series A co-led by First Round Capital and Bessemer Venture Partners, which took its total funding to $49 million.
The expansion itself is the expected move: a company that sells the drafting step for ground-based scanning adds the other way surveyors collect points. The more interesting document is the product page, because it spends as much space on exclusions as on features.
What It Does
Per Mach9’s aerial page, the pitch is “Turn aerial and drone LiDAR into CAD linework.” The workflow it describes:
- Classification on upload. The platform classifies the point cloud when it arrives, so a user does not have to pre-classify before extraction.
- AI-assisted extraction of planimetrics and overhead wires. Building outlines, roadway linework, tree lines and other top-down features come out as editable linework. AEC Magazine’s report, published the same week, adds kerbs, pavement edges and vegetation outlines to that list as 3D points and lines, and describes utility poles plus 3D catenary geometry extracted from suspended wires — feeding clearance analysis, make-ready workflows and pole-loading applications.
- Editing before export. Every line stays editable with CAD tools. AEC Magazine also reports surface modelling — ground surfaces, breaklines and contours, with vertex editing, surface regeneration and a deviation heatmap for checking the surface against the cloud.
- Inputs. Georeferenced LAS or LAZ, plus an optional GeoTIFF orthophoto. Minimum point density of 100 points per square metre, with 400 or more recommended.
- Sensor-agnostic. The page names GeoCue, YellowScan, Phoenix LiDAR, RIEGL, Leica and DJI among supported systems; AEC Magazine’s list also includes Trimble, Wingtra and Teledyne Geospatial.
- Exports. DXF, DWG, Shapefile, Geodatabase, CSV and Field-to-Finish CSV, with LandXML for surfaces per AEC Magazine.
Mach9’s claims for it are “centimeter-level accuracy, with Mach9 speed” and that firms run these projects “10x faster per technician.” Those are the company’s figures. AEC Magazine reports that the AI produces an initial draft within hours, which survey teams then check and edit before delivery — the same human-in-the-loop framing the company has used since the Series A.
The Exclusion List Is the Strategy
The aerial page states that Mach9 does not fly or collect data, does not plan flights, does not process trajectories or PPK, and does not register data. Users capture and process separately, then bring a georeferenced cloud in.
A product page that names what the product will not do is making a point about where its maker thinks the margin is, and here the point is legible. Trajectory processing, PPK and registration are steps with established incumbents and tight hardware pairings; flight planning belongs to the drone platforms. What is left — turning a classified cloud into linework a civil engineer can stamp — is a step that has stayed largely manual and is billed by the hour, and it is the step where corrected output accumulates into training data.
That is also consistent with the company’s own history. Mach9 was founded as Mach9 Robotics to build inspection hardware and dropped both the hardware and the word from its name; this publication’s account of the Series A was that the company had decided the scanner was not the bottleneck. Publishing an exclusion list is the same decision stated as product documentation. Read as analysis: a narrow surface is easier to make defensible, and far easier to make accurate, than a pipeline that owns every step from flight to deliverable.
The Customer Numbers, and Who Is Attached to Them
Two named customers give figures through AEC Magazine. Russell Hall of Langan — a senior survey project manager, per the magazine, working on Texas projects covering several thousand acres — reports that the software cut delivery time by around 40% and reduced the number of software packages needed; he is quoted saying “we’re able to build the surface natively inside, check the surface against the point cloud and the extraction.” Thomas Tucker, a survey project manager at Welch Comer, reports average savings of four to five hours per project. Mach9’s own aerial page quotes both men too — Hall on not being able to “compete on price with stereo mapping” on 3,000-acre projects before Mach9, and Tucker on the fact that “most of our LiDAR is aerial” — and adds Chris Gilling of McElhanney calling a fly-and-process workflow “orders of magnitude more efficient than traditional methods.”
These are customer testimonials supplied through the vendor and its trade press, with no methodology attached, and they should be read as such. They are still worth noting for one reason: the two figures are different in kind. Forty per cent off delivery time on a multi-thousand-acre job is a bid-level change — it moves what a firm can price. Four to five hours per project is a technician-level change. A product that produces both is being used at two quite different scales, which is more informative about adoption than either number alone.
The customer logos on the aerial page — NV5, Michael Baker, Woolpert, Olsson, HNTB, Bowman and Hanson — are large American engineering and geospatial firms rather than the small survey shops for which a saving of four or five hours a project would be the headline. Mach9 does not state pricing, a version number or a release date for the aerial capability on the page, and AEC Magazine’s report gives none either. For a product whose promise is that it lets a firm bid work it previously could not, pricing is the figure a buyer will want next. What the site offers instead is a banner for a live webinar, “Aerial LiDAR to AI-CAD”, dated 17 November.
