A contractor rarely loses money on the line item everyone is watching. The loss hides on page 40 of a spec book, in a single exclusion that turns an assumed cost into a change order nobody priced. FlowManual, a Y Combinator Summer 2026 company, was built on a plain observation: construction still runs on PDFs, and nobody has time to read all of them. So it built software that does.
The company describes itself as "the back office for construction contractors." In practice, that means an AI platform that ingests the documents a contractor drowns in - specifications, contracts, subcontractor quotes, invoices - and marks them up automatically. Scope items, prices, deadlines, and exclusions get highlighted directly on the PDF, with margin labels and clause references, before a human opens the file. It is less a chatbot than a colleague who pre-read the packet and left notes.
01 / The problemPaperwork is where projects are won and lost
Pre-construction is the stretch before a shovel touches dirt, when a contractor decides what to bid, buys out the job from subcontractors and suppliers, and signs the contract that governs everything after. It is enormously document-heavy and unforgiving. A missed scope change during the bid can turn a profitable job into a losing one. An invoice that quietly drifts from the agreed buyout price bleeds margin a few hundred dollars at a time.
The industry's default tools for catching this are a highlighter, a spreadsheet, and a late night. FlowManual's pitch is that those errors are both predictable and expensive, which makes them a good thing to hand to a machine.
"Cut the manual, repetitive work across bid, buyout, and contract review so contractors win more work and deliver at a higher standard."FlowManual's stated mission
02 / How it worksRead, compare, flag - by dollar impact
The product breaks into a handful of modules that map to the actual arc of pre-construction. Document intelligence handles the reading. A bid-versus-contract comparison puts a submitted bid next to a signed contract and ranks every scope addition, deletion, and price change by severity and dollar impact. Invoice verification checks incoming invoices against locked buyout prices and flags discrepancies line by line. Around those sit takeoff counts - an interactive pinning system for tallying items on drawings, cross-checked against the schedule - plus bid leveling to compare subcontractor bids on equal footing, and a purchasing view that tracks material price history with a 90-day outlook.
03 / The foundersTwo dropouts who like unglamorous problems
FlowManual was started in 2026 by David Shijoon Bae and Michael Lin, both of whom left Harvard to build it. Bae, the CEO, walked away from a mechanical engineering track. Lin, the CTO, is a published computer science researcher who, before this, built a video landing page that helped scale a product's revenue from $3,000 to $336,000 in monthly recurring revenue while cutting latency from 800 milliseconds to 100. That combination - one founder drawn to the physical trades, one obsessed with shipping fast software - shows up in the product's temperament.
"Built for the builders of America."Company tagline
The framing is deliberate, and a little defiant. Plenty of AI startups chase consumer flash. FlowManual chose bid leveling and invoice reconciliation, on the theory that the boring work is exactly where the dollars go missing. The company is based in San Francisco, was accepted into Y Combinator's Summer 2026 batch under partner Gustaf Alstromer, and raised a $130,000 seed in June 2026.
04 / The marketA crowded lane, entered on purpose
FlowManual is not alone. YC's Summer 2026 cohort leaned hard into construction and proptech, and a wave of AI bidding and procurement tools - Quotr, ContraVault, BuildOps among them - are circling the same back-office bottlenecks, alongside incumbent estimating and pre-construction software. What FlowManual is betting on is breadth stitched together: one platform that follows a document from bid to buyout to invoice, rather than a point tool for a single step.
Its stance on trust is part of the sell. Contractor bids and contracts are sensitive, so the company leans on AES-256 encryption at rest, SSO, MFA, and audit logging, and says its models do not train on user files without an explicit opt-in. The pitch to a skeptical superintendent is that the AI works alongside the team rather than replacing it.
05 / The businessFree to start, paid to scale
The model is freemium: free to start, no card required, with paid tiers and enterprise options - custom deployment, single sign-on, audit logging - for the larger general and MEP contractors that generate the most paperwork. That land-and-expand shape is common in software, but in a trade where new tools are met with suspicion, letting a contractor try it on one bid before paying is as much a distribution strategy as a pricing one.
Whether FlowManual becomes the system of record for construction's back office or a sharp feature others copy is the open question every early platform faces. For now, its wager is specific and testable: read the fine print, rank the surprises by how much they cost, and hand the builder a decision. In an industry that has run on highlighters for decades, that is a modest promise - and a measurable one. You can find the platform at flowmanual.com.