AI is going to make a lot of things a lot cheaper. But not all goods want to be cheap.

They’re called Veblen goods — products whose demand increases alongside their price. The canonical examples are handmade luxury items like Hermès bags. AI is likely to have limited effect on what they’ll cost to produce.

But the category also includes products that are most vulnerable to AI: the knowledge-economy work that Wall Street sells. They’ll soon be able to offer it more cheaply, but I doubt they will.

There’s been surprisingly little price competition for investment-banking advice over the years — you’ve never read about the Great Cola War of IPO underwriting because there wasn’t one. Clients don’t tend to buy more Goldman Sachs M&A advice when it’s cheaper. In part, Wall Street is immune to price pressure because the customers (corporate executives) are spending other people’s money (shareholders). But paying for top-shelf advice is a signal the same way that dropping $10,000 on a handbag is, and the stakes are a lot higher.

When the last crop of rainmakers spun out of big firms to launch their own boutiques in the late 2000s and early 2010s, they didn’t win business by undercutting incumbents. In fact, the Goldmans of the world welcomed the arrival of the Centerviews precisely because these newcomers charged a lot, propping up fees in the industry’s upper ranks.

“I hope not,” Lazard CEO Peter Orszag told me in April when I asked if his firm’s fees would shrink as AI automates grunt work. (I was thinking about all this as I read this Bloomberg story about Lazard cutting its fees to snake away a plum Venezuela assignment from a rival; white-glove service is rarely pitched as a red-tag special, even to broke governments.)

If good-enough advice becomes cheap and ubiquitous through AI, the premium on the prestige alternative goes up, not down. Fast fashion made Hermès more valuable as a signal precisely because the baseline got cheaper. Commoditization at the bottom of a market brightens the halo at the top.

The question now is whether investment banks can actively manage their businesses the way luxury houses do — resisting the temptation to expand downmarket and keeping the velvet rope up. There’s an argument to go the other way, to expand the pie of companies that might be willing to pay for the Goldman Sachs or Morgan Stanley imprimatur by pitching them a service that’s, say, 80% AI-generated and topped up by a human.

But nobody confuses Harvard Extension School classes with the real thing. I don’t expect Wall Street’s elite to stoop. They’ll cut whatever costs AI allows and let more of those fees drop to the bottom line.