The Barbell Is Already Here

Roman Pedan runs a hospitality company called Kasa, and he has a name for what he sees happening to the service economy. He calls it the "technology barbell effect." The largest platforms get stronger as AI cuts their coordination costs. The smallest specialists get capabilities they could never afford. And the mid-sized firm in between, with 15 to 50 properties in his world, gets squeezed. It maintains corporate overhead without the scale to fund differentiated technology. ^[1] Pedan named the pattern in his own industry, but he predicted it would spread to logistics, insurance, accounting, and healthcare. He was right. The barbell is no longer a forecast. It is already visible across the service economy, and the mechanism is simpler than most of the AI debate assumes. The mid-sized service firm is built on a staffing pyramid: a thin layer of senior partners or managers atop a wide base of junior professionals whose billable hours are the product. AI commoditizes exactly that base. The thing the mid-market sells is the thing AI makes cheap. Start at the top of the barbell. In May, OpenAI and Anthropic launched standalone implementation ventures, $4 billion and $1.5 billion respectively, to embed forward-deployed engineers directly inside mid-sized, private-equity-backed companies. The ventures are explicit about their target: they aim to "disrupt IT consulting," placing the model labs in direct competition with Tata Consultancy Services, Infosys, and Wipro. Blackstone and Bain Capital are anchor investors. The labs are not just building models anymore. They are vertically integrating into the services layer that represents more than three-quarters of U.S. GDP. ^[2] Now the bottom. In July, OpenAI launched a ChatGPT for Small Businesses program, describing the tool as a "force multiplier" that extends individual expertise for lean teams. It partnered with Shopify, Intuit, Dropbox, Slack, Atlassian, and Wix. ^[3] Fifty-eight percent of small businesses now use AI. ^[4] The result is not just survival at the bottom. It is active competition. A two-person brand-strategy firm called BrandGap.AI now compresses competitive positioning research from days to two or three minutes, at $199 per analysis. That is work a mid-sized branding agency would have staffed with a junior team and billed in the thousands. ^[5] Solitude AI, another small specialist, automates back-office workflows for logistics companies, cutting annual costs by up to $200,000 for small teams. Its client Legion Logistics shifted staff from billing administration to core operations. ^[6] Former McKinsey consultants are building AI tools to automate high-level consulting tasks, delivering with small teams what the pyramid staffing model of mid-sized consulting firms required dozens of junior analysts to produce. ^[7] The two ends of the barbell are moving simultaneously, and they are moving toward each other. The model labs are reaching down into services. The boutiques are reaching up into work that used to require a firm. The middle is being competed against from both directions at once. This is where the mechanism lives, and it is worth slowing down to see it clearly. Consider the mid-sized service firm: the consulting shop with a few hundred consultants, the business-process outsourcer with thousands of call-center agents, the branding agency with layers of junior strategists. All of them are built on a cost structure that AI dismantles. Revelio Labs data shows a 54 percent plunge in entry-level consultant hiring. Overall consulting job postings are down 26 percent year over year. McKinsey's headcount is down 11 percent. Bain lost 1,000 employees. AI is expected to impact 45 percent of consulting activities, and experts describe the shift as secular, not cyclical: something that will persist for a decade. ^[8] The same dynamic is playing out in business-process outsourcing. Microsoft cut 10,000 customer-service jobs and saved $750 million annually by replacing them with AI chatbots. Brinks halved its call center. Uber cut 10 percent. Commonwealth Bank shed hundreds. The mid-market outsourcing firms that built their businesses on those contracts, including Concentrix, Teleperformance, and TTEC, are declining as their clients transition to automated tools. Forrester estimates nearly half of all customer-service roles will be impacted by 2030. ^[9] And across the technology sector, the layoffs tell the same story. Nearly 50,000 AI-related jobs were cut in the first half of 2026 alone: 8,000 at Meta, 3,000 at Intuit (17 percent of its workforce), 2,250 at Acrisure, 1,000 at Wix. Pavel Durov, the founder of Telegram, put it plainly.

We're starting to see projects that used to require big teams now be accomplished by a single very talented person. — Pavel Durov

The mid-market is not being outcompeted on price or quality. It is being outcompeted on its own cost structure. The junior professional hours it sells are the very thing AI makes abundant and cheap. None of this means the barbell will empty the middle overnight. The counter-evidence is real and it deserves a straight look. AI operating costs still exceed human labor in many cases. Nvidia VP Bryan Catanzaro has said compute costs for his team "far exceed" employee costs. Uber exhausted its 2026 AI coding budget by April. An MIT study found AI automation economically viable in only 23 percent of vision-primary roles. ^[10] Off-the-shelf tools like Fireflies.ai and Otter.ai let some disciplined mid-sized firms adopt AI without owning compute or hiring technical teams. ^[11] Barclays finds that financial resilience predicts AI survival better than scale does: low debt, high productivity, strong cash generation. ^[12] And Silicon Valley firms are openly questioning whether the massive capex will earn its return. ^[13] But these are arguments about pace, not about structure. The barbell forms whether or not inference costs fall fast enough to hollow the middle tomorrow. The question is not whether the mid-sized service firm's staffing pyramid is the wrong cost structure for an economy where junior professional labor is commoditized. The question is how long the firms that occupy that structure can hold on. For the ones that do, survival will not look like a rebuilt pyramid. It will look like something closer to the boutique end of the barbell than the old middle. Aswath Damodaran, the NYU finance professor who has been warning of an AI correction, predicts what comes next.

My bet is that we hit peak AI a few months ago, and that there will be more consolidation and correction in the months ahead — Aswath Damodaran

The mid-sized firms that adapt will not stay mid-sized. They will shrink into specialists or get absorbed. The middle does not reconstitute.