AI Took the Grunt Work Your Juniors Learned On. Nobody Decided Who Teaches Them Now.

Every senior strategist at your firm learned the job the same way: by building the deck wrong. The v1 that came back covered in markup. The model that didn't reconcile, circled in red. The "this slide is about us, not the client" comment that stung for a week and changed how they wrote for a decade. The grunt work was never just grunt work — it was the apprenticeship, and the markup loop between junior draft and senior correction is how a firm's judgment actually moved from one generation to the next.

In 2026, AI does the v1. Which means the loop that produced your best people is quietly being switched off — and almost nobody has decided what replaces it.

The bottom rung is gone

The data stopped being anecdotal this year. Harvard research finds that at companies adopting generative AI, entry-level hiring has fallen roughly 80% per quarter since 2023, even as senior headcount at the same firms keeps growing. A British Standards Institution global survey found 39% of business leaders have already reduced or cut entry-level roles because of AI, and 43% expect to do so in 2026. Stanford economists tracking payroll data put a number on who absorbs it: a 13% relative decline in employment for 22-to-25-year-olds in the most AI-exposed occupations, while older workers in the same jobs held steady.

Professional services sits at the sharp end, because the junior work — research synthesis, first-draft documents, slide assembly — is exactly the high-volume, structured, document-heavy profile that generative AI handles best. Between 2018 and 2024, the share of consulting jobs requiring three years of experience or less dropped from 41% to 26%. The associate who used to spend two years building proposal decks before touching strategy now largely doesn't exist. The deck builds itself; the associate was the line item that paid for it.

Read as a cost decision, this is rational. Read as a talent pipeline, it's a firm eating its seed corn. The partners of 2036 were supposed to be the analysts of 2026 — and the analysts of 2026 were supposed to learn the job on work that no longer reaches them.

The deck was never just a deck

Here's what the cost decision misses: the low-value work and the high-value learning were the same activity.

A firm's real standards are tacit. What "client-ready" means at your shop. Which claims need a source and which can stand. How much confidence the numbers have to earn before they go on a page. What your firm's voice sounds like when it's right and how it curdles when a junior imitates it badly. None of that lives in the brand guidelines. It lives in senior heads, and for a hundred years the transmission mechanism was correction: the junior produces, the senior marks up, the junior internalizes, repeat until the judgment transfers.

The curse of knowledge cuts both ways here. Seniors can't simply tell juniors the standard, because expertise is precisely the stuff you can no longer articulate — you see the weak slide instantly but can't fully explain the seeing. The markup loop worked because it didn't require articulation. It required reps. Take away the reps and you haven't made the apprenticeship more efficient; you've cancelled it.

Supervision without standards is a rubber stamp

The official plan for the entry-level role is supervision: 96% of HR leaders expect entry-level jobs to evolve into "AI supervision positions" within five years, and a 2026 LinkedIn Economic Graph review found about one in four entry-level consulting and finance postings now require AI fluency — including "ability to validate AI output" — up from fewer than one in twenty two years ago. The junior of 2026 is hired not to produce the draft but to check the machine's.

There's a hole in the middle of that job description. Validating output requires a standard to validate against, and the standard is the exact thing juniors used to acquire through the drafting they no longer do. We're asking the least experienced person in the building to quality-control a system that produces work above their own unaided level, against criteria nobody ever taught them. The predictable result already has a name: Stanford and BetterUp researchers found around 40% of workers received AI "workslop" from colleagues in the past month — polished-looking, substance-light work that took nearly two hours per instance to sort out — and recipients rated the senders as less capable, less creative, and less trustworthy afterwards. Workslop is what supervision without standards ships.

For a client-facing firm, that verdict doesn't land on a colleague. It lands on the client reading your proposal.

Client documents are where the gap surfaces first

The proposal, the pitch, the board pack — these are the places a firm's tacit standards become visible to someone paying for them. And the review capacity that used to catch the gap is thinning at both ends: fewer juniors learning the standard on the way up, and seniors who now personally re-review everything because there's no trained middle layer to delegate to. The partner who used to see the third draft now sees the first — and the first draft came from a model.

Sarah's strategy boutique feels it as an evidence-standard problem: the AI draft asserts confidently, and the junior reviewing it has never been marked up hard enough to know which assertions her firm would never let out the door. Alex's growth agency feels it in the numbers — a junior who never built the reporting deck by hand doesn't flinch at a CAC figure that contradicts the appendix. Maria's creative shop feels it as voice drift the newest hire literally cannot see, because she's three months in and the AI's version of the agency's voice is the only version she's ever known. Raj's ops consultancy and Julia's comms firm feel it as the compliance and nuance misses that used to get caught by someone on their fifth hundred document.

None of these are AI failures, exactly. They're transmission failures. The machine did what it does; the human check the client is paying for never got built, because the person doing the checking never got taught.

Write the markup loop down

The answer isn't to refuse the leverage — the firms that reinstate manual deck assembly as a training program will lose on cost to the firms that don't, and juniors know busywork when they see it. The answer is to take the thing that made the apprenticeship work — senior judgment, applied to real work, with the reasoning attached — and make it explicit, so it scales without a partner standing over every shoulder.

That's what a knowledge-grounded review layer actually is: the markup loop, written down. In Lurio, the firm's standards stop being tacit — your past winning proposals, your evidence bar, your brand system, your voice — and become the knowledge that review agents check every page against. Data Integrity flags the number that doesn't reconcile with the appendix. Brand Compliance catches the voice drift the three-month hire can't see yet. Audience Fit asks whether the slide is about the client or about you — the same question the red pen used to ask. And because every critique is cited back to the firm's own knowledge, the junior doesn't just get a correction; they get the reasoning, on every document, which is the part of the apprenticeship that actually taught.

Custom review agents close the loop from the other side: when a senior articulates a standard once — "we never state projected ROI without naming the assumption" — it becomes a check that runs on every page the firm sends from then on. Senior judgment stops dying in senior heads. The partner's review shifts from catching what the junior missed to judging what only a partner can judge, and nothing ships without a human's sign-off — which keeps the final call exactly where the client believes it is.

The firms that navigate the next five years won't be the ones that kept juniors doing grunt work, or the ones that quietly stopped training anyone. They'll be the ones that noticed the apprenticeship was hiding inside the workflow — and rebuilt it into the tools before the last generation that learned the old way retires.

The ladder lost its bottom rung. Build the standard into the work, and every document becomes a rung.

— The Lurio Team