Bobyard said in a Business Wire release on 29 October that its platform now carries a bid from construction drawings through a priced estimate and a client-ready proposal to a submittal package, in one product. The company’s own framing of the sequence, in its press update the same day, is that it has moved “beyond the takeoff.”
The reason that is worth more than a feature-list story is the choice of destination. A submittal is not a step in winning work. It is paperwork that only exists because the work was won.
What a Submittal Is, and Why It Is the Odd One Out
The FAQ on Bobyard’s own submittals post defines it plainly: “A submittal is a formal document a contractor sends for review and approval after winning the bid, confirming that planned materials and methods comply with project specifications. Product data, like manufacturer cut sheets, is the most common type and is required on almost every project.”
Everything else Bobyard sells — takeoff, cost database, estimate, proposal — is pre-award. It is work done on jobs a contractor may not get, which is exactly why automating it has such a legible return: the same estimator can price more bids, and the bids that are lost cost less to have made. A submittal is post-award. The contractor is already being paid. Automating it does not buy more shots on goal; it buys hours back on a job already on the books.
Bobyard’s submittals feature post, dated 1 October, reports what its beta participants said the job costs them today: “In a survey with beta feedback participants, most said any given submittal took them 2-5 hours to prepare, with some at 10 hours or more.” The company says the product cuts that roughly in half. Two anonymised lines from the same beta programme are published on the page, one of them: “For irrigation and soil submittals, it was roughly twice as fast as the way we did it before.”
Note the sequencing, because the two documents do different jobs. The feature itself is dated 1 October on Bobyard’s own blog; the 29 October release is the announcement of the connected end-to-end workflow that submittals completes.
How It Works
The mechanism is a matching problem rather than a drawing problem, and Bobyard describes it as such. The descriptions below are split across two of the company’s documents — the submittals post and the submittals product page — and are quoted from whichever carries the wording. Materials come across from the takeoff as submittal items with descriptions pre-filled from the drawing legend. Product catalogues are uploaded once and split into individual products, with manufacturer, model and size extracted into a reusable, company-wide library. An assistant the company calls Bob GPT then “searches your library and the web to find the right product data for every item on your list, then highlights the model numbers.”
The control step is the one to note: “Just like a takeoff, every item shows ranked matches with confidence scores, so you stay in control of what goes to your GC.” The output is a single branded PDF with cover page and table of contents, and a resubmittal rebuilds only the flagged items.
The same human gate sits on the takeoff side. Bobyard’s homepage FAQ answers “Who does the takeoff, the AI or my estimator?” with “Both. The AI does the first pass across the drawings. Your estimator reviews each detection, using the confidence score on it, before it is locked into the estimate, so the estimator stays in control of scope and final pricing.”
The Numbers, and Whose They Are
The release’s traction figures are Bobyard’s own and are reported here as such: customers cut takeoff time by an average of 65 percent, and submit three to five times more bids per estimator. Michael Ding, founder and chief executive, is quoted in the release on why the company began where it did:
“We started with takeoff because if you cannot read the drawings, you cannot automate anything that depends on them.”
The release’s industry framing cites the Associated Builders and Contractors’ estimate that construction needs roughly 349,000 additional workers in 2026, and says “the Federal Reserve Bank of Richmond found that real output per worker in construction in 2020 was more than 30 percent below 1970 levels.” Both are the release’s citations rather than this publication’s findings.
Named customers appear on Bobyard’s own pages rather than in the release, which names none. Marty Grunder, founder of Grunder Landscaping, is quoted on the submittals page saying that takeoffs built around materials let a team “grab linear feet, area, and volume all in one pass,” and that “we’re talking cutting takeoff time in half on real jobs.” Gustavo Fonseca, Project + Takeoff Engineer at New Path Landscape, is quoted on the same page: “Before, I’d spend my entire day doing a takeoff when now I can spend maybe like an hour doing the takeoff, and I can spend the rest of the day working on submittals and getting pricing.”
Trade by Trade, Not Market by Market
Bobyard’s expansion pattern is the most informative thing about the company, and it is documented on its own site rather than inferred. The homepage FAQ states: “As of Q4 2026, Bobyard supports six trades: landscaping, electrical, plumbing, mechanical, interior finishes, and paving. Each has its own trade-specific takeoff model.”
The company’s account of how it got there, from the press update: landscaping first, then electrical, plumbing, paving and finishes, with its own computer-vision models taught “trade by trade.” The 29 October release says that “with today’s news Bobyard brings AI from takeoff to submittals for the following additional trades: drywall, insulation, doors, and windows.”
The reading that follows is this publication’s own. That is a deliberately slow way to grow a software company and a defensible one. An electrical drawing and a landscape plan do not share a symbol vocabulary, and a model trained to count irrigation heads has no idea what a panel schedule is. The trade-specific model is the moat and the bottleneck simultaneously: each new trade is a dataset and a training run rather than a configuration screen, which caps the rate of expansion and also means a general-purpose competitor cannot arrive at parity by adding a dropdown.
Bobyard was founded in 2023 by Ding and is based in San Francisco. It announced a $35 million Series A on 9 December 2025, led by 8VC with participation from Pear VC, Primary Venture Partners, Tishman Speyer, RXR and Caffeinated Capital. That post also named drywall, electrical, HVAC, plumbing and framing as the trades it intended to expand into with the money.
Where This Sits
The submittal has turned up repeatedly in this newsroom’s coverage this year, and the interesting thing is that each company reaches it from a different starting position. In date order:
- Newforma’s Vojo, announced at Newforma World 2026 in May, aims at submittal review — the reviewer’s side of the document, on the design team’s desk rather than the trade contractor’s.
- Buildertrend put AI into submittals and AIA-style progress billing on 29 September — a residential platform assembling the document and billing apparatus that commercial work runs on.
- Procore unveiled a submittal agent at Groundbreak on 21 October, which ENR described as similar to an existing agent that can update a submittal in every place it is referenced across a set of construction drawings, and reported as live.
- Bobyard, on 29 October, comes at it from the takeoff — which means it arrives at the submittal already holding the quantities and the materials the submittal has to describe.
That last point is the structural argument for Bobyard’s version, and it is this publication’s reading rather than a claim from the company. A submittal tool that starts from a blank project has to be told what materials are on the job. A submittal tool downstream of a takeoff already knows, because the legend it read to count the items is the same legend that names them. The connected workflow is not a bundling exercise; it is the configuration in which the submittal step inherits its inputs for free.
The adjacent pattern is worth noting too. Preconstruction has drawn a steady run of small rounds this autumn — in date order, Quotr’s $4m seed, announced 30 September; Constrik’s €350k pre-seed, reported in the Spanish startup press in mid-October; Tyce’s $2m for reading the bid package on 19 October; and PataBid’s $2m pre-seed for trade takeoffs on 20 October — and every one of those is aimed at the pre-award half of the job. Bobyard is pushing the other way, into the half where the contract already exists.
What to Watch
The checkable claim is the trade count. Six trades as of Q4 2026, each with its own model, is a figure the company dates itself in its homepage FAQ; whether it is eight by the spring will say more about the company’s model-training throughput than any customer metric.
The second is whether submittals stay free of the thing that makes post-award automation hard. Pre-award, a wrong number costs a contractor a bid. Post-award, a wrong cut sheet goes to a general contractor as a formal representation that the materials comply with the specification, and getting it wrong has consequences further down the chain than a lost bid. Bobyard’s answer is the confidence score and the human confirmation step — the same answer the takeoff product gives. It is the right answer, and it is also the one that determines how much of the promised time actually comes back.
