Do we still need the analyst who builds slides if the AI builds better slides faster?

The shift

Producing the artifact — the slide, the model, the first-pass analysis that is the junior analyst's entire output — goes from scarce (a trained person's hours) to abundant and near-free. And AI is frequently better at the surface craft than a first-year: cleaner layout, tighter prose, faster iteration, fewer formatting errors. What stays scarce is structuring a problem nobody has framed yet, sitting in front of a client and being trusted, owning the recommendation when the business acts on it, and the years of reps that turn a junior into someone with judgment.

The axioms

1. Turning scattered inputs into a clean, client-ready deck or model takes analyst-hours, so a junior's core output — the thing they're staffed and paid for — is production.
2. A polished, well-structured artifact is a reliable proxy for the quality of the thinking behind it: if the slide is tight and the model ties out, the analysis is probably sound.
3. Juniors do production grunt work for years as apprenticeship — building the deck by hand is how they absorb how a problem gets structured, what the numbers mean, and what "good" looks like before anyone trusts them with a client.
4. The analyst is scoped to production and kept out of the room because they lack the context, standing, and accountability to frame the problem, face the client, or own the call — those belong to seniors.
5. Someone junior must exist to do the volume of production a senior can't, which is why teams are staffed with a wide base of analysts under each partner.

Invalid axioms

1. A junior's core output is production — the deck and the model. Drafting slides, laying out a first-pass analysis, and building a first-cut model from provided inputs is exactly the plausible-first-draft generation AI is strongest at, and it now does the surface craft faster and often cleaner than a first-year. The habit-trap: firms still staff analyst classes sized for production capacity and measure a junior's day by artifacts shipped, when producing the artifact is no longer the scarce or valuable part.
2. A polished artifact is a proxy for sound thinking. This heuristic worked because polish used to be expensive — only someone who had done the work could produce a tight deck, so tightness signaled the work was done. AI severs that link: it produces a flawless-looking slide with a wrong or unexamined number underneath just as easily as a right one. The habit-trap: reviewers still read professional formatting as evidence of rigor and skim the artifacts that "look done," which is precisely where confidently-wrong output hides.
3. You need a wide base of juniors to do the production volume. The staffing shape assumed production was labor-bound. When one analyst with good tools produces what took a team, the headcount justification for the base thins. The habit-trap: firms keep hiring the analyst class at its old size and shape, filling the gap with make-work, rather than resizing around the parts of the role that survive.

Unchanged axioms

1. Structuring an ambiguous problem is scarce. Deciding what the question actually is, what to analyze, and which cut of the data would change the client's mind is judgment under novel ambiguity — there's no pattern to match because the useful framing is specific to this client's situation and politics. AI produces an answer to the question you hand it; deciding the question is the harder, still-human part. A junior who can do this is doing something the tool can't, and it's the thing worth training toward.
2. Client-facing judgment and trust are scarce. Reading whether the client is ready to hear a conclusion, which stakeholder will kill it, and how to land it are learned in the room and don't transfer to a vendor tool. This was never what the junior was paid for — but it's what the role was supposed to grow into.
3. Accountability for the recommendation stays scarce. Someone answerable when the business acts on the advice and it's wrong — a career and a firm's reputation on the line — is exactly what a model can't carry. A better slide changes nothing here; the deck was never the thing being trusted.
4. Judgment is still built by reps, not conferred. No one becomes a partner by reading a memo about judgment; it comes from doing the work, being wrong, and being corrected, many times. The reps still hold as the mechanism. What no longer holds is that hand-building decks is the rep that produces them — see NEW.

New axioms

1. When production is free, nobody has defined what the analyst does instead. If the artifact takes an afternoon of prompting, the junior's day has to be about framing the problem, interrogating what the AI produced, and pushing toward the right answer — work that used to be reserved for people two levels up. Whether a first-year can do that work, and how you hire and train for it, is unsolved; the role is being emptied of its old content faster than a new definition is arriving.
2. The training ground is gone, and the reps that built judgment ran through the production that disappeared. The apprenticeship worked because building the model by hand forced a junior to touch every assumption. Automate the building and you keep the judgment requirement (STILL HOLDS #4) but delete the mechanism that produced it. If the reps now have to come from framing and verifying rather than producing, no one has designed that path — and the gap doesn't show until the current juniors are supposed to be seniors and aren't ready.
3. Telling a good slide from a right answer is now the core skill, and it's harder than producing either. When a plausible, beautifully-formatted deck is free, the scarce act is judging whether it's correct for this client — verifying the number, catching the framework applied to the wrong context, noticing what the confident output quietly assumed. This is a more advanced skill than deck-building, and we're now asking it of the most junior person on the team, or of a reviewer who's inclined to trust anything that looks finished. Who verifies, and how they learned to, is unsettled — and this is a fast-moving call: as models get more reliable and better at multi-step research, the volume of plausible-but-unverified output rises faster than verification habits adapt.

Where it breaks

The polish-as-proxy heuristic (invalid) collides with telling a good slide from a right answer (new): reviewers trained to read a tight deck as a sound one are now most exposed exactly where the artifact looks most finished, and the person expected to catch the error is the junior whose old job — building the deck — was the thing that used to teach them what to catch.

A second collision: firms still staff and measure juniors by production volume (invalid) while the reps that turned production into partner-grade judgment have quietly vanished (new). The role is being paid for the output AI now makes free, at the moment the firm most needs it to be the training ground for framing and verification it hasn't rebuilt.

ai-axioms.vercel.app/axioms/management/consulting-junior-analyst-ai-axioms