Tyce Technologies announced a $2 million pre-seed round on 19 October, led, in the company’s own wording, by the Business Development Bank of Canada’s Thrive Venture Fund and the MaRS Investment Accelerator Fund. The Toronto company’s own announcement says “$2 million ($2.8 million CAD) in pre-seed funding” without labelling the first figure’s currency; the Construction Broadsheet reads it as US dollars. The company says the money goes to its engineering team, its core AI model capabilities, and expansion into the North American market.
At that size the round would normally be a line in a weekly roundup. Two things make it worth more than a line. The first is what the company says the model is trained on. The second is that a named general contractor has put a specific number next to the result — which is the thing this newsroom has spent most of 2026 asking construction AI companies for, and rarely gets at this stage.
The private-corpus pitch
Tyce’s framing is that the asset is not an industry-wide model but each contractor’s own archive. The announcement’s load-bearing sentence is plain enough: “Tyce gives each contractor a private AI system configured with its own project data.” The product indexes bids, proposals, RFPs and contracts, then applies that history to new pursuits — reviewing drawings and specifications, flagging scope gaps and risky contract terms, supporting estimating and trade-bid management, and drafting proposals. The company’s own site describes it as “the AI platform that contextualizes your historical project data, to win projects and manage preconstruction risk,” and says Tyce “indexes your bids, proposals, RFPs, and contracts, then puts that institutional knowledge to work on every new pursuit.”
The Construction Broadsheet’s report, published 21 October, fills in the mechanics. Drawings, specifications, addenda and trade bids are pulled into one project model a team can query. Each flag cites the sheet, spec section or page it came from. Documents come in through Procore and Autodesk Construction Cloud integrations, and scope sheets, bid levelling and risk summaries export to Excel. It is sold as an annual subscription tiered by company size and usage rather than per project, and the Broadsheet reports that no historical data is required to start, with setup and training usually taking a few weeks.
That last detail is the commercial answer to the obvious objection: a system whose value compounds with your archive still has to be useful on the day it is switched on, before much of that archive is in it. Chief executive and co-founder Lylia Djait-Paulien is direct about why the archive stays shut in, telling the Broadsheet that a contractor’s pricing and bidding strategy “is their competitive advantage,” and that customer data is not used to train models for other firms.
“The things that get missed in preconstruction are what turn into change orders and claims later.”
— Lylia Djait-Paulien, CEO and co-founder, Tyce
Djait-Paulien is joined by Anuar Yeraliyev as chief technology officer and co-founder. The company’s announcement does not give further background on either founder, and we have not independently established their prior employers.
The number, and whose number it is
Eastern Construction, an Ontario general contractor, has rolled out Tyce across its preconstruction teams and reports cutting bidding and project-management time by at least 40%. President and chief executive Dean Walker is quoted in Tyce’s announcement: “With Tyce, tasks that took our preconstruction team five hours now take 15 minutes.”
Two things about that. It is Eastern’s figure, relayed through the company it is praising, and it should be read that way — a customer’s account of its own experience, not an audited result. But Eastern is not a logo of convenience. Its leadership page lists Walker as “President & C.E.O.” at the head of a seven-person executive team, and elsewhere the company describes itself as one of Canada’s employee-owned construction companies and cites 75 years in business. A chief executive of a general contractor of that vintage attaching his name and a number to an early-stage product is a different quality of evidence from an anonymous “a top-10 ENR contractor.”
The Broadsheet notes that Tyce has not published accuracy figures or win-rate data. That is the gap that matters. A five-hours-to-fifteen-minutes claim is a throughput claim; it says nothing about whether the scope gaps the system flags are the scope gaps that would have cost money, or whether the ones it misses are the expensive ones. Preconstruction AI is graded on recall against a set of omissions nobody catalogued at the time, which is why the whole category keeps publishing speed numbers and not catch-rate numbers. Our look at the two takeoff benchmarks that landed eleven weeks apart on 19 October found that even where a yardstick exists, it was authored and scored by a vendor selling the product being measured. Scope review is harder still, because there is no agreed ground truth to score against at all.
The investors
Both leads are Canadian public-purpose funds rather than generalist venture firms, which shapes what the round signals.
BDC Capital’s Thrive Venture Fund invests specifically in women-led Canadian technology companies from pre-seed to Series B — its page states the mandate as investing “in women-led technology companies to accelerate their growth and transform today’s markets.” Managing partner Mona Minhas says of Tyce that “Lylia and her team combine deep industry insight with a practical application of AI.”
The MaRS Investment Accelerator Fund deploys roughly $10 million a year across pre-seed and seed-stage Ontario companies and has invested more than $100 million since 2008, by its own account, across more than 200 ventures. William Ma, named its managing director in April, calls Tyce “positioned to define the next era of construction management software.”
A $2 million round from two funds with public-purpose mandates is not capital that forces a sprint for market share. For a product whose defensibility claim is depth inside individual contractors’ archives rather than breadth across the market, that is arguably the right shape of capital — a point of analysis, not a fact the investors have stated.
Where this sits
Preconstruction is one of the most crowded corners of construction AI this newsroom covers, and the cheques in it run from €350,000 to $22 million. LightTable raised $22 million to catch errors before ground is broken. Quotr took $4 million at seed for estimating and procurement, and PataBid’s $2 million pre-seed came after nine years of building without venture money. Constrik raised €350,000 to do a version of this for Spanish builders. Tyce’s differentiation is not the workflow — takeoff, levelling, scope review, proposal drafting is becoming the standard feature list — but the insistence that the corpus is the customer’s and stays that way.
Tyce expects the United States to become its largest market, the Broadsheet reports, citing infrastructure spending in Texas, California and New York. That ambition is the real test of the round, because a product that configures itself to one contractor’s archive and takes a few weeks of setup per customer is a high-touch sale — and a $2 million pre-seed does not buy many hands to make it with.
What to watch
Whether a second customer puts a number to the result. Tyce’s site already carries named testimonials from Dominic Ho at Tier Construction and Tim Ropchan at MaisonMTL, but neither quantifies anything — Eastern’s is still the only figure in public. And whether anyone publishes a catch rate rather than a time saving, which would tell us whether the flags are worth the subscription. Until then the honest summary is that a small, well-sponsored round has bought a credible product a year of runway and one unusually specific public reference.
