My final-year thesis asked a question that sounds like it answers itself: would moving freight from lorries to trains through the Pyrenees cut emissions? It would. I measured it. And that turned out to be the least useful thing in the whole project. The corridor is Zaragoza – Canfranc-Estación – Pau. Mountain freight, an international railway line that has spent decades mostly dormant, and a road crossing that 65% of the professionals I surveyed rate as high-risk. The thesis had three stated objectives: estimate the current carbon footprint of road freight on that axis, calculate the reduction potential of a partial modal shift, and compare the energy efficiency of the two modes. All three are answerable with published emission factors and arithmetic. I answered them, and they point exactly where you would expect them to point. The full thesis, as submitted for my Business Management degree, is available in PDF (in Spanish): Análisis del transporte ferroviario como alternativa sostenible al tráfico pesado en los Pirineos — 60 pages, July 2025. The part that taught me something was the part that was not in the objectives: asking the people who actually move the freight whether they would use it. The survey looked like a green light I surveyed hauliers and freight operators — roughly 70% drivers, 30% intermediaries — on a five-point scale, then followed up with interviews. The headline numbers are encouraging. 72.5% scored 4 or 5 on adopting a combined road-rail model. 60% believed rail could be tariff-competitive. 70% agreed it would reduce their exposure on the most dangerous stretches. Read only that and you conclude the market is waiting. Then you reach the item asking whether they would accept a longer transit time in exchange for a lower carbon footprint. Mean score: 3.15 out of 5. Only 40% agreed. The same people who overwhelmingly want the modal shift will not pay for it in time. That is not a contradiction. It is a specification, and I had been treating it as an attitude problem. The conditions were the actual finding In the interviews, the conditional half of that 72.5% turned into numbers. Not preferences — thresholds, stated without hesitation, the way people state something they have already worked out: → Punctuality at or above 95%. Below that, one operator explained, buffer stock comes back and penalties at destination eat the saving. → Door-to-door cost no more than 10-15% above all-road. → A fallback plan inside 6 hours if the window is missed. → Ground time under 90 minutes per unit, with the slot confirmed the day before. → Two services a day. One makes it possible; two make it plannable. → And volume: 200-300 tonnes a week aggregated, or around 250 tonnes a month per account. Below that, terminal handling costs swallow the fuel saving whole. A driver gave me the physical version of the same thing: on a bad day, the climb costs him an extra 25 to 35 litres. He knows precisely what the road costs him. He also knows precisely what he would need in exchange to give it up. The result I had not predicted I ran a chi-square test of independence to check whether a respondent's role — driver or intermediary — predicted their acceptance of the model. It did not. No significant relationship. I had expected the two groups to want different things. Drivers carry fatigue and risk; intermediaries carry margin and reliability. They were arriving at the same answer from opposite directions, and the answer was not the environmental one. What I took from it The environmental case was never the blocker. It was never the trigger either. It was the reason the question was worth studying, and it was irrelevant to whether the change would happen. What gates adoption is an operational guarantee that somebody is willing to sign: a punctuality figure, a cost ceiling, a recovery time. Until those exist in a contract, an emissions comparison is a true statement about the world that changes nothing in it. I finished that thesis and moved into AI, and I keep running into the same shape. The model benchmarks better. The comparison is honest. The pilot still does not ship, because nobody will write a number next to it that they would then be held to. The technical question tends to get settled early. The question of what happens when it fails, and who absorbs that, is the one that decides. If you are trying to get something adopted and you are still arguing that it is better, you have probably already won that argument — and lost the one that matters.