There’s a trap catching AI startups in construction right now, and OpenSpace CEO Jeevan Kalanithi named it on the show. Leadership hands down an “AI strategy,” buyers sign deals to satisfy it, and founders mistake that for product-market fit. Then they over-hire, dilute, and hit a wall when nobody actually uses the thing. Across industries, 95 percent of enterprise AI pilots deliver zero measurable return. Here’s how to tell a real signal from an expensive one.
Jeevan Kalanithi has run OpenSpace since 2017, which by construction-tech standards makes him an elder. So when he called this “a dangerous time” for AI companies, it wasn’t doom-posting. It was pattern recognition.
The danger he’s pointing at is sneaky because it looks like success. Boards and executives across the industry have decided they need an AI strategy, and they need to be seen having one. That creates a wave of buyers who will sign something, almost anything, to check the box. For a founder watching deals close faster than ever, it feels like the market finally woke up. Jeevan’s warning is that a lot of that demand is a mirage, and building your company on it is how you die slowly.
Because a signature bought to satisfy a mandate isn’t the same as a customer who can’t put your product down. Tell the two apart and you build something real. Confuse them, and you scale a business with no floor under it.
A deal signed to satisfy a “we need AI” mandate looks identical to product-market fit on the invoice. It’s the opposite on the usage graph. Watch the usage graph.
The mandate that feels like traction
Why “we need an AI strategy” is a trap for founders
Here’s the mechanism Jeevan laid out. A leadership team decides the company needs AI. That pressure rolls downhill until someone’s job is to go buy some. They find a vendor, sign a deal, and everyone up the chain gets to say the box is ticked. The founder on the other side of that deal sees revenue land and reads it as proof the product is winning. It usually isn’t. It’s proof the buyer had a mandate.
The industry data backs up how wide that gap runs. A Dodge study found that 87 percent of contractors expect AI to meaningfully change construction, but only 19 percent have actually adapted their workflows to use it. That 68-point gap is the space where mandate-driven deals live: bought with enthusiasm, never wired into how work gets done. Across industries, roughly 95 percent of enterprise AI pilots deliver zero measurable return, and about 85 percent of those failures trace back to messy data.
The trap springs when the founder treats that early revenue as a green light. They hire against it, raise against it, and dilute to fund the growth. Then the mandate cools, the unused tool gets cut in the next budget review, and the company is left carrying a cost base built on demand that was never really there. Jeevan calls the endpoint churn death, and it’s quiet, because the invoices looked great right up until they didn’t.
- The mandate. Leadership says “get an AI strategy,” so someone is sent to buy AI, any AI.
- The signature. A deal closes fast. The founder reads speed as demand and revenue as fit.
- The delusion. Hire, raise, dilute. Then the mandate fades, usage was never there, and the whole thing unwinds.
The only metric Jeevan trusts
Usage over revenue, even on his own products
What makes this credible is that Jeevan holds his own company to it. On OpenSpace’s early-access AI agents, he told his team he doesn’t much care about the revenue yet. The number he watches is whether people keep coming back and using the thing without being nudged. Revenue from a mandate is a lagging, flattering metric. Repeat usage is the truth, and it shows up early.
It’s the tractor test, really. Nobody buys a tractor to satisfy a “farming strategy.” They buy it because the work is impossible without it, and then they use it every single day. That’s the bar. If your customers would notice and complain the moment your product vanished, you have something. If they’d shrug, you have a line item waiting to be cut, no matter what the contract says.
This is also why racing on feature count is a distraction. A conservative buyer can’t absorb a new workflow every two weeks, so shipping faster just widens the gap between what’s sold and what’s used. As Dustin put it on the episode, the market’s already drowning in near-identical takeoff tools. Another one, launched louder, doesn’t change whether anyone opens it twice.
I don’t care about the revenue yet. I care whether they actually keep using it.Jeevan Kalanithi, CEO, OpenSpace
The construction tech moves that matter, before everyone else clocks them
Deals, tech, and honest takes for people who build, fund, and sell into the industry. One email, no fluff.
Join 3000+ ReadersHow to tell real from mandated
What to watch instead of the contract
For founders, the discipline is simple to say and hard to hold: chase usage depth, not signed pilots. A deal without daily use is churn on a delay. Instrument retention from day one, and be honest when a logo looks great but the login graph flatlines a month in. That’s not a customer, it’s a countdown.
For investors, the tell is in the revenue quality. Mandate-driven revenue is fragile and should be discounted, not celebrated. Ask where the demand came from, whether anyone would fight to keep the product, and what usage looks like ninety days after signature. A cap table built on box-ticking deals is a cap table built on sand.
And for buyers, the fix is the oldest one in the book: problem-pull, not tech-push. The contractors who actually get value start from a pain they already feel, prove it on one job, and only then scale it. On a live site, if your shiny tool becomes just another thing to manage, the team quietly forgets it. Buy the thing that solves a real problem, not the thing that satisfies a slide in the board deck.
Mandated deal vs real product-market fit
Same invoice, opposite futures
On paper, a box-ticking deal and a can’t-live-without-it deal look the same. The difference shows up everywhere except the contract. Here’s the split.
| Signal | Looks like traction (mandate) | Actually is traction (fit) |
|---|---|---|
| Why they signed | Leadership said “get an AI strategy” | A named pain they couldn’t ignore |
| Who champions it | A committee ticking a box | An operator who’d fight to keep it |
| Usage after 90 days | A spike at launch, then drift | Daily, unprompted, deepening |
| The revenue | Real but lagging and fragile | Grows with usage, expands on its own |
| Renewal risk | Cut when the mandate cools | They renew before you ask |
It’s when a company’s leadership decides it needs an AI strategy, so someone is sent to buy AI to satisfy that. The resulting deals look like product demand, but they’re really box-ticking. Founders who mistake them for product-market fit over-hire and dilute against revenue that vanishes when the mandate cools. (Source)
Not on its own. A signature bought to satisfy a mandate can sit next to zero real usage. The proof is whether people keep using the product without being nudged. Revenue lags; retention leads. (Source)
Repeat usage and retention. Jeevan Kalanithi says he watches whether people keep coming back to OpenSpace’s AI agents before he cares about the revenue. If usage flatlines after launch, the revenue is on borrowed time. (Source)
Across industries, about 95 percent of enterprise AI pilots deliver zero measurable ROI, and roughly 85 percent of those failures trace back to poor data quality. In construction, undocumented workflows make it worse, because an agent can’t replicate a process that lives in someone’s head. (Source)
Start from a pain your team already feels, not from “we need AI.” Prove value on a single project, then scale. If the tool adds friction on a live site, it gets quietly dropped no matter what it cost. Problem-pull beats tech-push every time. (Source)
In aggregate, yes: around 74 percent of construction organizations have minimal or no AI capability, and only about 27 percent of AEC professionals currently use it. But the market is split, not uniformly slow. Some segments, like trade contractors, are well ahead. (Source)
Related Articles
“You Can Only Go As Fast As Your Customers”: The Quiet Brake on ConTech AI
The Construction Giants’ Buying Playbook, From Inside Vinci and Kajima
