The Ticket to Nowhere

Many CEOs burn money on AI solutions that create more problems than they fix. Meanwhile, the skeptics dismiss AI that creates value today.

There's one set of people who win big together. First are the frontier-captains building at the edge who know what customers want and what AI can do. Second are the frontier-passengers who know their destination cold — a payroll audit, a building portfolio, a border — and buy from the frontier-captains who can chart the route.

Frontier-passengers don't have a gap in intelligence. They are the most capable people in their own domains, with twenty years of judgment a vendor will never match. But they still need frontier-captains to help them cross an ever-changing ocean of AI without getting sunk.

The money is made in the partnership between frontier-captains and frontier-passengers.

Anduril Didn't Wait for the Charter

In 2017, the U.S. government knew it had a border problem. But it could not write the requirements for a technology to fix it.

Nothing in the federal acquisition catalog in 2017 described an autonomous sensor tower that ran computer vision on-board and handed an operator one clean picture of incursions instead of nine noisy screens. Back then, defense procurement almost always started from a government-authored specification. And defense buyers — like most buyers — usually cannot specify technology they don't know.

Anduril stopped waiting for the government to order the right charter. It built autonomous towers, now sold as Sentry, and the Lattice software behind them on its own venture money, then put them in the field for Border Patrol to test in 2018. Customs and Border Protection didn't buy a specification. It bought a solution that its staff could watch and confirm worked much better than the status quo — and by 2020 had made the towers a program of record.

But the towers worked because Anduril read the hazards, not just the destination. Legacy border surveillance meant an agent sitting in front of screens, chasing false alarms, driving out to check a sensor that had tripped on a jackrabbit. The gain was detection. The hazard was the labor and the noise — and that is the half most vendors never bother to price.

Here's a key piece of Anduril's asymmetric financial returns. The Lattice software it wrote to run surveillance towers became the orchestration layer for everything from counter-drone systems to autonomous aircraft and mission command across its defense systems. The instruments built to sail one route became the charts for the fleet.

Anduril was marked at roughly $90 billion in its latest tender. The towers were the first voyage. Lattice was a ship Anduril could outfit for even more valuable trips later on.

Nobody Charts the Route Alone

Here is where most captains fail, and it isn't a failure of ambition.

A real nautical chart is mostly hazards. Depths, wrecks, rocks, shoals, the water you do not enter. The destination is a single dot. The rest of the sheet is warnings. A captain seeing only the destination point will navigate straight into shipwrecks.

The captain knows the frontier. The passenger knows the hazards. Neither one can draw the route alone.

Only the passenger knows where the rocks are in their own business — which failed audit costs them a license, which process eats forty hours a week, which customer leaves if response time slips one more time. That knowledge is not in your training data and it is not on their website. It is the most valuable thing in the room, and nobody is pitching it to you.

Too many AI founders don't ask buyers how their solution could make them fail. Instead, founders fall in love with the frontier and chart a beautiful route that buyers don't want. They are flummoxed when their pilot contracts fail, and blame the buyers for not seeing the frontier.

Knowing what your passenger wants to avoid is worth as much as knowing what they want to gain. Daniel Kahneman, whom I studied under at Princeton, showed empirically in Prospect Theory, his most cited paper with Amos Tversky, that people will take more risks to avoid a loss than they will to make a gain. The thing your passenger wants to avoid is already costing them money — which means the buyer often already has a budget for it.

Four Tickets — But Only One Wins

Cross two questions and you get a map I now use on every AI deal. Does adopting AI here earn the buyer's money or burn it? And does the buyer believe that adopting AI will make them money?

The Winning Ticket is where frontier AI founders need to focus today. The passenger is a seeker who already believes AI will make them money, and they're right. The budget is already there, the proof comes fast, and every passenger who lands safely becomes the demonstration that sells the next ticket.

The Empty Seat is the future market. The ride was real, it would have paid, and the person who most needed it stood on the dock — because nobody sold them a ticket they could believe in. Nobody is a permanent skeptic; they are one demonstration away from being a passenger. The frontier founders selling winning tickets today will be best positioned to fill these seats tomorrow.

The Ticket to Nowhere is the expensive one, and I want to be clear about whose fault it is. The easy story blames an overexcited buyer. The truer story is a captain who never asked what the passenger was trying to avoid, charted to the most impressive destination on the board, and sold a fare against it. The pilot program never converts to a sale. The passenger is stranded, as this MIT report describes, and reluctant to board again.

How Trebellar Wrote a Winning Ticket

Trebellar began discussions with Meta's commercial real estate (CRE) team, which is staffed with operators who are exceptionally good at running buildings. In fact, Meta's team runs its CRE operations at the highest level of the industry.

Trebellar's founders, a former Salesforce VP and a former Google AI leader, didn't just pitch Meta on a new AI platform. They learned all the reasons frontier AI was failing, and might fail, Meta's CRE team. After months of working together, Meta signed a large contract and has been shifting other budget to Trebellar because the team keeps shipping AI that adds value. Meta's head of global workplace services now says the platform is "quickly becoming a key part of how our team thinks about real estate decisions."

That detail is the whole thesis in miniature. Budget moves from incumbent vendors to frontier AI teams when they deliver value. Trebellar gets paid for avoiding the hazards, not just for seeing the horizon.

When Trebellar wins with frontier seekers like Meta, it's easier to convince skeptics to buy the ticket Trebellar already knows wins.

The second customer is the test. If customer two costs the same to serve as customer one, you weren't selling a winning ticket. You were running a consultancy with a software multiple attached.

Out of Imagination

At GTC in March, CNBC's Jim Cramer asked Jensen Huang why companies keep cutting staff over AI instead of doing more with more. His answer was one of the bluntest — and most correct — things I've read:

"Because you're out of imagination. For companies with imagination, you will do more with more."

That is the gap, priced by the person with the best view of it. The skeptic who can't picture the route stands on the dock and misses a ride that would have earned them a lot of money. The seeker who books a route that was never there burns the fare and tells the market AI doesn't work. Both lose money. Neither is stupid.

The winning captain sells the ticket that adds value, and refuses to sell the one that sinks the passenger. And the captain can't do it alone. They need a passenger willing to show them where the rocks are. If not, they sink together.