Consulting is a delicate contract: endure two challenging, formative years – and in return, get a golden ticket to anywhere. Firms like McKinsey tout themselves as the “CEO factory”, and boast they’re “not surprised” to be consistently named the best place for future leaders.

The skills they promise to build – synthesis, sharp analysis, crisp communication, client-readiness, hypothesis-driven thinking – have enticed every generation’s top graduates. Get an offer from a place like this, and everything else will fall into place: about as clear a guarantee of future success as you could get fresh out of a bachelors. These firms spent decades marketing themselves as production houses of excellence, and until recently, they were.

But that value proposition no longer holds in the age of AI. Analysts are “either using AI for their own efficiency or being told to”, Zain Mobarik, a former consultant, put it to me. Another former consultant, speaking on the condition of anonymity, recalled a year-one analyst on her team asking if she could help him get the work done. The conventional approach would be to carve time for coaching, a “give me two minutes and I’ll work it through with you,” the way that all of us learned. He corrected her: “No, could you just send me the prompts to put into AI?” She did. His output gleamed well beyond his year of experience.

“The business analyst programme was the best possible training ground in existence 10 years ago,” Romil Depala, former analyst now at a London-based PE fund, told me. In the years before AI, the role mostly fulfilled that promise: stuck right into challenging client work, taught to turn around clean financial models in a couple of days and polish decks to crisply communicate our razor-sharp analysis. Analysts would learn by owning their own corner of knowledge, getting in the weeds deeply enough to “crack” something, then present it in front of a client. Your progression, and your learning, was contingent on that ownership.

Now, this execution lies mainly with an AI. With the likes of BCG’s Deckster, Bain’s Sage or McKinsey’s Lilli – GenAI for internal knowledge management that can answer 500,000+ prompts each month – entry-level roles (for those lucky enough to still get them) have essentially become factchecking. There’s a question mark over the whole apprenticeship model – it’s unclear how analysts will build skills like we used to, while the skills we did learn (developing your own framework, creating work plans and creative ideation) are redundant when AI can replace them all anyway.

Firm leaders frame this as getting junior colleagues up the ladder faster. “They’re just going to be doing things that are more valuable to our clients,” one senior partner told Bloomberg. But this is just spin. Firms are trimming their workforces and freezing salaries, while the people getting promoted are AI-fluent and high-churn – the ones that grease the wheels of this new business model.

Clients, driven by fear and opportunity, are desperate to infuse AI across their entire business, the kind of multi-year transformations that require “deep implementation expertise”. Enter the service-as-software model: a race to partner with frontier AI companies as a matter of survival. For junior staff on the ground, this looks like arduous, incremental work, with compressed delivery time and fixed fees tied to deliverables rather than time inputs – an unenriching AI slog that leaves little room to learn.

So the kind of people these firms are recruiting is already changing. A final-round interview might now include using a firm’s internal AI to produce output. While publicly following the same enticing playbook as the AI labs – championing EQ and creativity – firms are clearly looking for churners over thinkers, analysts who can run multiple agents at once to pluck IP from existing assets as fast as possible.

The role that used to offer unparalleled growth has essentially become an accelerated plug-and-play. Sure, there’s some value judgment and good prompting required, but how is this supposed to attract the CEOs of the future? The classic skills that would put juniors at the forefront of any industry have been reduced to “a workforce of 40,000 humans (complemented by) 20,000 agents”.

Then, there’s the exit options – the whole point of the consultancy contract. Endure, then enter into an elite pool from which to get plucked out by another equally elite firm: proof, as Mobarik put it, that “you can do things quickly and with a high level of execution,” crafting “an inner circle of high-performers”.

But with white-collar work at risk across the board, even this elite cycle is slowing – “a pool of 300 analysts going for the same five jobs”. Consulting’s entry power used to be one of the only routes into such firms, but as roles dry up, the pathway has broken.

And the famed consulting-to-startup ramp makes less sense than ever: not only has AI eroded the grind and competencies these exits depended on, but spending two years incubating means missing the rush entirely. The skills that now matter are split: the technical ones, such as deploying a full stack, are best learned by building AI itself; the ever-more-coveted human ones, such as empathy and tolerance, give public service an edge.

So, where does this leave the sharpest graduates of today? The machine that mass-produced capable generalists is being dismantled quicker than anything replacing it – and nobody, including the firms, knows what comes next. Where the next generation of leaders will be made is now up for grabs.

• Alice Lassman is an economist who writes The Intimacy Economy, a Substack and forthcoming book on the economics of connection, care and relationships