Contech Funding: ConstructAI Pitched 97 Funds, Then Closed

ConstructAI shut down in September 2026. Founder Pedro Neira wrote the post-mortem himself, with no AI tools, and put numbers on every cause.

Pedro Neira wrote his own post-mortem. He says he used no AI tools, only his memories and his records.

In September 2026, ConstructAI ran out of funds and shut down. Neira, its co-founder and CEO, put the numbers in writing.

$275K raised. 97 funds pitched. 15 reached IC or data room. 6 invested. $325K in hard commitments at shutdown. $1.6M in “follower interest” at shutdown.

The story is a case study in contech funding. Every figure here is Neira’s own, and MAIN has not verified them.

What ConstructAI Built

ConstructAI was an AI platform that turned structural 2D drawings into native Revit models, with concrete and rebar detailing, Neira says.

He describes it as 5x to 10x faster than manual modeling. It aimed to add modeling capacity with the existing workforce and save 80% of modeling cost.

The company began in May 2025 at the Latitud Fellowship in San Francisco. Neira says he spoke with 35 construction companies in the last two weeks, and all 35 shared the problem.

He then recruited Fernando Maytorena, an architect and BIM developer, as co-founder and CTO. The team peaked at five people.

Two Pivots in 12 Months

Neira ran pilots in Mexico, Ecuador and Peru. Five were signed, and four were paid.

In October 2025 he pivoted from five trades in Latin America to structures only, in the United States. He calls the U.S. market 20 to 40 times bigger.

In January 2026 came a second pivot, from takeoff estimation to BIM modeling. Neira lists the pivots as a decision, not a mistake.

“In hindsight, none of the pivots were wrong. I firmly believe we made those pivots for the right reasons, and wouldn’t change them, but they surely came at a cost: each pivot means that at some level you restart and change direction, and it eats away a couple of months of cash and traction.”

Two Viral Videos

In March 2026, Maytorena posted a video of ConstructAI automating 1,000 architectural elements from a PDF to Revit in 10 minutes.

Neira reports 75K post impressions and 25K video views, all organic. A waiting list of more than 170 followed.

Autodesk, Amazon and JLL asked for live demos. But the product was still a prototype on Maytorena’s laptop.

The beta went live on April 17, after seven weeks near burnout. The full launch came in July.

A second video showed more than 22,000 structural elements with rebar detail going from PDF to Revit in eight hours. It drew 100K impressions and 30K video views.

What Worked

“None of this erases what we built. Two viral videos that proved out a repeatable acquisition motion. 200+ cumulative pipeline leads, several with Fortune 500 names, several with signed NDAs to try our product (Amazon, Whiting-Turner, JLL). A product that shipped and worked: columns and piles automation, client-verified on video at 77% and 82% time savings. The BIM modeling time savings were real, and at the pace we were building each new element, the product would’ve been finished in 4-6 months more. None of it was enough to outrun the cash clock, but all of it was real.”

The Funding Gap Nobody Talks About

Neira’s primary cause is about who could lead the round. He says contech is unlike the rounds he had raised before.

Generalist funds, he says, cannot underwrite the technical risk. He compares contech to biotech and deeptech.

“Generalists wanted a Contech fund to lead; Contech funds wanted traction first.”

He adds a rule for founders. “If you don’t have a vertically focused VC that can lead your round, you won’t get funds from generalist/agnostic VCs.”

The Sequence Problem

No funding → no team → no product → no traction.
No product → no funding → no team → no traction.

It runs in both directions, Neira writes. “I bet on the wrong direction.”

Sales Cycles Ran Long

Neira expected enterprise sales cycles of three to four months. They ran six to twelve.

Signing an NDA took over a month. Pilot terms and pricing took two to three more, and legal approval followed.

A security audit usually added another month. “Longer sales cycles affect traction and cashflow. Traction affects fundraising,” he writes.

What the Numbers Actually Say

Metric	Value
Months of operation (payroll to shutdown)	17
Team size at peak	5
Total capital raised	$275K
Funds pitched	97
Reached IC or data-room stage	15
Investors in ConstructAI	6 (3 angels, 3 early-stage VCs)
Hard commitments at shutdown	$325K
Follower interest at shutdown	$1.6M
Monthly burn at shutdown	$25K
Paid contracts	4 ($70K total)
Active pipeline leads at shutdown	45
Fortune 500 NDAs signed	4
Waitlist before beta	170
Organic reach of two launch videos	175K impressions
Product completeness at shutdown	60%
Estimated time to sellable product	5-7 months
Client-verified time savings	5.8x faster BIM modeling

Neira gives two estimates for finishing the product. The table says five to seven months. His “What worked” section says four to six.

A Cooler Market for Contech

Neira was raising into a selective market. Cemex Ventures counted $1.85 billion in contech venture funding across 68 transactions in the first quarter of 2026.

That was down 33% from a year earlier. Cemex called it the weakest quarter since the first quarter of 2024.

Cemex read the drop as normalization and greater selectivity, not a collapse. About 80% of deals were at pre-seed through Series A, and 60% involved AI-enabled products.

Strategic investors took part in 35% of transactions, according to Cemex.

MAIN cannot say where Neira’s fundraising fell in that cycle. The trend is context for his account, not proof of its cause.

Miami Investors Describe a Similar Gap

Miami investors describe a related problem. Mark Volchek, founding partner of Las Olas Venture Capital, told MAIN that local depth thins as companies scale.

“There are fewer locally based investors writing larger Series A and growth checks… so many of the strongest companies still need outside capital,” he said.

Ron Tarro, president of New World Angels, said the same pressure runs statewide. “You’re not competing for capital locally. You’re competing statewide and nationally.”

Joe Roos, CIO of ZFO, framed fundraising as a managed process. He listed “a targeted list of sector focused funds, consistent outreach, disciplined follow-up, a current data room and enough runway to preserve negotiating leverage.”

That maps onto Neira’s finding. He pitched 97 funds before learning that generalists would not lead.

Neira also left Latin America for the U.S. market. Roos told MAIN that Miami’s edge is proximity to customers in sectors he named.

He named “real estate, fintech, hospitality, wealth management, logistics, healthcare and/or Latin American commerce.” Construction technology was not on his list.

The Lesson for Founders Everywhere

“What I’d do differently. Sequence product development so that an MVP can generate enough traction with whatever cash you have available. Traction will close your lead investor; follower investors are 10x easier to close. Be more pessimistic with B2B/enterprise sales cycles, and always structure the round around the lead, not around agnostic or generalist funds.”

What He Would Not Change

“We tackled a significantly difficult technological problem to solve a humongous inefficiency with very limited resources. Yes, we were optimistic about our chances, but isn’t that the job of an entrepreneur? Isn’t our job to aim to solve problems that weren’t solved before with little or no resources? I believe it is, and I’ll keep trying, applying what we’ve learned here, of course, to avoid making the same mistakes.”

Neira also plans a fully AI-generated post-mortem in two weeks. He wants to compare the two accounts.

ConstructAI ran out of cash before it ran out of road. Neira says he will keep trying.

Note from the Editor: We wish Pedro Neira the best. A founder this creative, this honest, and this willing to tackle hard problems doesn't stay down long. He'll land on his feet. The next one will be better.

Pedro Neira is a 4x startup founder. His full post-mortem is available on LinkedIn. ConstructAI was based in Austin, Texas.

Sources: