This page, a companion to the HECC Manifesto, is where the Layer Split gets specific — for your role. The AI wave isn’t compressing your job title — it’s compressing part of it, and lifting the rest. Where The Elevated Roles names the new roles the wave is creating, this page splits the role you already hold.
The manifesto makes the case for the move, and the Layer Split is the first step. Most people can’t see the line between the two layers of their job — so they stay in the execution layer, get faster at the work that’s disappearing, and never move up.
Each entry draws that line for a single role. It names the execution layer plainly, names the judgment layer that stays human, and hands you the exercise that makes your own split visible. Selective, not exhaustive — built role by role. If your role isn’t here yet, request it here ([form URL]).
Marketing / Content Manager
For anyone whose job is getting the right message to the right people — and who’s watched AI start doing the producing.
The compression. The execution of marketing has collapsed in under two years. First-draft copy, blog posts, ten social variations from one idea, email sequences, image options, the summary of last quarter’s numbers — work that used to fill your week now takes an afternoon. AI doesn’t decide what to say. It just says it, fast, in volume, on command.
What stays yours. The decision of what’s worth saying, to whom, and whether the thing in front of you is any good. AI will hand you forty headlines. Knowing which one is true to the brand — and which thirty-nine are plausible noise — is judgment, and it gets more valuable as the producing gets cheaper. So does the call on what not to publish. So does owning the result when the campaign ships.
The split. Take one week of your work and write down every task — every deliverable, every meeting, every decision you made. Then sort each line into one of two columns: did this produce something, or did it decide something? Do it by hand, in a blank doc. Don’t hand the sorting to AI — the writing is what forces you to see where your week actually goes.
The move. Start handing the producing to the machine — drafts, variants, repurposing, the first cut of anything — and move the hours you get back into the deciding: the positioning, the brief, the audience, the review that catches what’s off before it ships. You don’t need permission or a budget to begin. Open a doc, write down this week, draw the line down the middle. That’s the first step, and it costs nothing.
Marketing was one of the first white-collar roles where the execution layer fell away in plain sight. The people who move up aren’t the ones producing faster — they’re the ones who decide.
Operations Manager
For anyone who keeps the business running — and spends half the week pulling together numbers instead of acting on them.
The compression. The reporting layer of operations is going first. The weekly status deck, the reconciled spreadsheet, the inventory summary, the first draft of an SOP, the dashboard nobody had time to read — AI now assembles in minutes what used to eat your Monday. It pulls the numbers, writes the update, flags the outlier. The compiling that filled your calendar is becoming a command.
What stays yours. What to do about the outlier. AI can tell you the line went down; it can’t decide whether that’s a supplier problem, a staffing problem, or noise — and it can’t carry the call when you’re wrong. The tradeoffs between cost, speed, and quality, the decision of which process to fix first, the read on whether a team is stretched or slacking — that’s judgment, and it gets more valuable the moment the reporting stops eating your time.
The split. Take one week and write down every task — every report you pulled, every meeting, every call you made. Then sort each line into two columns: did this compile something, or did it decide something? Do it by hand. The point isn’t a tidy list — it’s watching how much of your week went to assembling information you barely had time to act on.
The move. Hand the compiling to the machine — the reports, the reconciliations, the monitoring, the first draft of any process doc — and move the recovered hours into the deciding: which process is costing you most, where the real bottleneck is, what to change next. You don’t need a new system or a budget line. Open a doc, write down what is going on this week, draw the line. That’s the first step, and it costs nothing.
Operations is the clearest case of a role drowning in execution that was never the actual job. The manager who moves up is the one who stops compiling the numbers and starts acting on them.
Software Developer
For anyone who writes code for a living — and has watched the model finish their function before they did.
The compression. Writing code is the part going fastest. Autocomplete that finishes the function, boilerplate generated from a comment, unit tests written on request, the bug with a known shape fixed in seconds, the documentation that writes itself. On real-world engineering tasks — actual issues pulled from real codebases — AI went from solving almost none to solving the majority in under two years. The typing was never the job. Now everyone can see it.
What stays yours. What to build and why. AI will write the function you describe; it won’t tell you the feature shouldn’t exist, that the architecture won’t hold at scale, or that the clever fix just opened a security hole. Naming the right abstraction, deciding what not to build, reading why the system is actually slow, owning the code when it’s running in production at 3 a.m. — that’s judgment, and it gets scarcer and more valuable as the typing gets free.
The split. Take one week and write down what you actually did — every ticket, every pull request, every decision in a review. Then sort each line: was this producing code, or deciding something about the code? Do it by hand. The surprise for most developers isn’t how much they wrote — it’s how much of the real value was in the deciding they barely counted as work.
The move. Let the machine write the first draft of everything — functions, tests, boilerplate, the obvious fix — and move your hours up into the design: what to build, how it should fit together, where it’ll break, whether it should ship. You don’t need a new title or permission. Open a doc, write down what you worked on last week, draw the line between the code you produced and the calls you made. That’s the first step, and it costs nothing.
The developers who move up aren’t the ones who type fastest with an AI copilot — they’re the ones who decide what’s worth building and whether it’s any good. The machine writes the code. It doesn’t own it.
Financial Analyst / Accountant
For anyone whose work is numbers — and who spends more time assembling them than interpreting them.
The compression. The mechanical layer of finance is compressing fast. Reconciliations, transaction categorization, variance tables, the model built from last quarter’s template, the first-draft commentary on the numbers, the formatted statement — work that filled the close is now minutes of machine time. AI pulls the actuals, ties out the accounts, and drafts the explanation. The assembling that defined the junior years is evaporating.
What stays yours. Whether the number is right, and what it means. AI will reconcile to the penny and still miss that the assumption underneath is wrong. Judging materiality, choosing the forecast assumptions, reading what the variance actually signals about the business, deciding what to tell leadership and what to flag to the auditor, putting your name on the position — that’s judgment, and it’s the entire reason the role exists once the arithmetic is free.
The split. Take one month-end or one week and write down every task — every reconciliation, every model, every report, every call. Then sort each: did this assemble a number, or judge one? Do it by hand. Most analysts discover their week was spent producing the inputs to decisions other people made — and that the path up runs through the judging, not the assembling.
The move. Hand the assembling to the machine — reconciliations, first-draft models, variance tables, standard commentary — and move the hours into the interpretation: which assumptions hold, what the numbers mean for the next decision, where the risk actually sits. You don’t need new software or sign-off. Open a doc, write down last close, draw the line between the numbers you produced and the ones you judged. That’s the first step, and it costs nothing.
Finance has always paid most for judgment and least for data entry — Wave 5 just widened the gap. The analyst who moves up is the one who stops tying out the numbers and starts telling the business what they mean.
Project Manager
For anyone who’s accountable for getting things delivered — and spends the week updating the plan instead of steering it.
The compression. The administrative layer of project management is going first. Status updates, meeting notes, the schedule that needs re-baselining, the follow-up email, the risk log nobody reads, the weekly report — AI now captures, summarizes, and drafts in minutes. It transcribes the standup, updates the tracker, writes the stakeholder update. The coordination overhead that ate your week is becoming background work.
What stays yours. The decisions about what matters. AI can update the timeline; it can’t decide which deadline to defend and which to let slip, read that two leads are quietly at war, or carry the call when you cut scope. Prioritizing under real constraint, managing the people, reading what’s genuinely at risk versus what’s noise, owning the delivery when it’s late — that’s judgment, and it’s what the role was always actually for.
The split. Take one week and write down everything you did — every update, every meeting, every nudge, every decision. Then sort each line: did this report on the work, or steer it? Do it by hand. The pattern almost every PM finds is the same — most of the week went to keeping people informed, and the steering happened in the margins.
The move. Hand the reporting to the machine — notes, updates, tracker maintenance, the status report — and move the hours into the steering: the priorities, the tradeoffs, the people, the risks that actually threaten delivery. You don’t need a new tool or authority you don’t have. Open a doc, write down last week, draw the line between reporting and steering. That’s the first step, and it costs nothing.
Project management was always supposed to be judgment under constraint — the admin just buried it. The PM who moves up is the one who stops narrating the work and starts directing it.
Recruiter / HR
For anyone whose job is people — and who spends the day screening and scheduling instead of deciding who’s a fit.
The compression. The processing layer of recruiting and HR is compressing fast. Resume screening, candidate sourcing, interview scheduling, the first-draft job description, the summarized interview notes, the outreach template, the policy doc no one wanted to write — AI now drafts and sorts in minutes. It ranks the pile, books the calls, writes the rejection. The high-volume processing that filled the day is becoming a query.
What stays yours. Who’s actually a fit, and the conversations that decide it. AI will rank a hundred résumés and miss the one person who’d transform the team — and it can’t read the room in an interview, handle a layoff with dignity, or carry the call on a borderline hire. Judging fit beyond the keywords, reading what a manager actually needs versus what they wrote down, the hard human conversations, owning a hire that goes wrong — that’s judgment, and it’s the part of the work that was always human.
The split. Take one week and write down every task — every screen, every schedule, every conversation, every decision. Then sort each line: did this process people, or decide who fits? Do it by hand. Most recruiters and HR pros find the same thing — the week filled up with processing, and the real call on who fits got whatever time was left.
The move. Hand the processing to the machine — sourcing, first-pass screening, scheduling, draft descriptions and templates — and move the hours into the judgment: the fit calls, the manager conversations, the interviews that matter, the people decisions only a human should own. You don’t need a new platform or approval. Open a doc, write down what you worked on last week, draw the line between processing and judging. That’s the first step, and it costs nothing.
HR was always meant to be about people — the paperwork just crowded it out. The recruiter who moves up is the one who hands the processing to the machine and spends their judgment where it counts: on the humans.