The New Org Chart: Who's Good at What

Anya is staring at the headcount plan for next year. Twelve roles, ranked, costed, one-line justifications. Her COO put it together over the weekend and it’s the cleanest version of this document the firm has ever produced. She reads down the list — senior associate, senior associate, ops analyst, ops analyst, junior on the research desk, project coordinator, knowledge manager, billing analyst — and stops.

Seven of these don’t make sense in the firm she’s been describing for six months.

The list is right. The model behind it is wrong. She’s looking at a 2019 org chart with 2026 numbers attached. Her COO wrote it in good faith because the shape of an org chart is the deepest assumption a firm has about itself, and nobody’s told him out loud that the shape is changing.

The shape is changing.

Three buckets, every responsibility

Pick any role in your firm. Pick yours. Write down everything that role actually does in a week — not the JD, the actual things. Reconcile the WIP. Brief the partner. Chase the client on missing data. Sit through the staffing meeting. Read the room.

Every activity lands in one of three buckets.

Bucket 1 — deterministic execution. Rules, math, hand-offs, audit trails. Reconcile WIP. Generate the engagement letter from a template. Apply the standard discount schedule. Anything where, given the inputs, the right output is mechanically derivable.

Bucket 2 — judgment-with-context. Classify, summarize, decide under ambiguity, draft. Pull the three things that matter from a client’s annual report. Triage the inbox. Draft the partner brief. Anything where the answer is defensible but not deterministic — where a senior person, given the same inputs, would converge on roughly the same call.

Bucket 3 — human-only. Relationships, taste, accountability, exception-handling, strategy under genuine uncertainty. The conversation with the client where you read the silence and decide not to push the upsell. The partner meeting where four people disagree about a new vertical. The decision to fire a client.

Every responsibility maps to one of three. Not two. Not “kind of one and kind of another.” Force the choice. The exercise feels uncomfortable for an hour, then very clarifying.

Bucket 1 — what computers were always good at

Bucket 1 has been hireable to a machine for fifty years. OCR read handwritten addresses in 1965. Spreadsheets did the math in 1979. ERPs routed transactions in the 1990s. Every wave of business software has been a deeper bite into bucket 1.

So why is your firm still staffing it with humans?

Because the judgment steps in between needed people, and the cheapest way to wire judgment and execution together was to put both in the same skull. Your ops analyst doesn’t just reconcile the WIP — she notices the Mercer numbers look weird this month, escalates, and holds the invoice until the partner confirms. Reconciling is bucket 1. Noticing-and-escalating is bucket 2. One human did both because separating them cost more than fusing them.

That’s no longer a reason. Judgment is hireable now too. A flow reconciles, notices the anomaly, drafts the escalation, routes it to the right partner. The human shows up at the one point where the judgment isn’t bucket 2 — where it’s actually bucket 3. Everything else runs without her.

Most firms ask “what can we automate in bucket 1?” Wrong question — bucket 1 has been automatable for decades. The right question: now that bucket 2 is hireable, which bucket-1 work were we still doing only because it was glued to bucket 2 in someone’s JD? In most firms, most of it.

Bucket 2 — the new bucket

Bucket 2 is what changed.

Classify, summarize, decide under ambiguity, draft. These used to require a person — inputs messy, rules soft, output had to read like a human had thought about it. Triaging inbound RFPs. Drafting an engagement memo. Pulling implications from an earnings call.

For thirty years, bucket 2 was the moat of professional services — the part associates spent five years learning, the reason a senior associate is worth more than three juniors.

Bucket 2 is now hireable.

Notice the word — hireable. Not solved, not automated. Hireable, the same way bucket 1 became hireable to a payroll system in 1985. Put a bounded specialist on a bucket-2 task, get output good enough to send to a senior reviewer, who corrects, signs off, and moves on. The senior stays in the loop. The five-year apprenticeship that used to be the prerequisite to be in the loop does not.

This is where every founder gets the model wrong. You don’t need AGI to staff bucket 2. You need bounded specialists, not a polymath — small models tuned to specific tasks, wired together by an orchestrator. The capability has been here for eighteen months. It costs roughly the second coffee your associate buys on Tuesday.

What changes when bucket 2 is hireable? The pyramid stops needing the bottom. Not the bottom people — the bottom function. The work juniors did for their first three years just got hireable. Juniors stop being apprentices and start being something else.

Bucket 3 — what’s actually human-only

Shorter list than people think. More valuable than people think.

Relationships. The thirty-second exchange where you read the room and decide to eat the overrun. The dinner where you find out — not from the agenda — the CFO is leaving. None of this is content. It’s contact. A substrate cannot have it on your behalf.

Taste. The judgment that this engagement is on the firm’s line and that one isn’t, even though the rate card says they’re equivalent. The call that the deck reads cheap even though every box is checked. Taste is what’s left when the rules run out — which means it’s what your firm sells, even when it pretends to sell something else.

Accountability. Somebody has to answer the phone at 9pm when the engagement is on fire. Somebody has to put their name on the deliverable. Accountability is not a task — it’s a standing. You can’t delegate it to a flow. You can barely delegate it to another human.

Exception-handling. The substrate works for the modal case. Reality keeps producing the non-modal one. The client whose data doesn’t fit the schema. The matter crossing three jurisdictions the routing rules didn’t anticipate. Bucket 3 is where the firm meets the world as it actually is.

Strategy under genuine uncertainty. Not the staffing call where the answer is on the rate card. The Berlin office. The new vertical. The decision to fire the client whose fees are 14% of revenue. Data is incomplete by definition, analysis can’t resolve, and the act of committing is the value-add. The substrate lays out trade-offs. It cannot make the call.

Five categories. In a firm of fifty, maybe six FTEs’ worth of activity.

The trap is that bucket 3 feels like the firm — which makes it easy to overstate. Most of what your senior people do is bucket 2 dressed up as bucket 3, because for thirty years it had to live in a senior-person skull to get done at all. Clock the calendar honestly: a senior associate’s bucket-3 hours are 8 to 12 a week. The rest is plumbing and judgment-with-context, both of which now have somewhere else to live.

Redrawing the org around the buckets

Not a re-org chart with the same boxes renamed. Most “AI re-orgs” are exactly that — take the existing JDs, add “with AI assistance,” move everyone up a level, wave at the savings. Repainting the same building.

The boxes change shape. Roles dissolve, recombine, new ones appear.

Roles that dissolve. Most of “operations support.” The work that existed because senior people couldn’t be expected to do their own plumbing. The ops analyst reconciling the WIP. The project coordinator chasing people who haven’t responded. The knowledge manager curating the shared drive. None were bad jobs — they were necessary given the substrate. No longer necessary.

Roles that compress. The pyramid’s bottom three layers compress into one. The three-year apprenticeship that produced partners who could read a P&L because they’d built fifty has to be redesigned — the work that was the apprenticeship has moved. Juniors are still hired, but for bucket 3 readiness, not bucket 1 throughput. Shadowing senior people from week one, not assembling decks for them.

Roles that appear. The new role is the flow designer — the person who maps the deterministic, judgment, and human-only steps of a recurring process and wires them together. Which model handles which step. Where human review points sit. What the audit trail looks like. What happens when the substrate gets it wrong. Part business analyst, part product manager, part ops engineer. They used to build the workbook everyone keyed numbers into. Now they build the substrate everyone runs through.

A flow designer is worth two senior associates. Most firms don’t have one. By 2027, you can’t run a serious firm without three.

The other shape change: the partner-to-staff ratio gets higher, not lower. Bucket-3 work scales with seniority; bucket-1-and-2 scales with the substrate. Anya’s firm of thirty-five, in the new shape, is fifteen partners, eight bucket-3 specialists, four flow designers, and a substrate. Twenty-seven total. Same revenue or higher.

That math is why this matters. It’s also why her COO’s twelve-hire plan needs a different document.

What this means for hiring next year

Back to Anya’s twelve.

Two senior associates — “support the partner on Tier 1 engagements.” 70% bucket 2. A flow with one bucket-3 reviewer at the top covers it. Cut.

Two ops analysts — “WIP reconciliation, billing support, partner reporting.” 95% bucket 1. The flow her commercial director prototyped last quarter covers it for a single seat license. Cut.

A research-desk junior — pure bucket 2. Cut, redirected. What she needs: one bucket-3-track junior, shadowing senior partners from day one, learning taste rather than process. One, not three.

A project coordinator — pure bucket 1. Cut. A knowledge manager — looks like taste, actually plumbing. Cut. A billing analyst — bucket 1 with the occasional dispute, already covered by the partners. Cut.

That’s seven roles she doesn’t need. The two she keeps — a senior partner-track hire and a domain specialist for the new vertical — are bucket-3 hires. Both stay.

The role that didn’t exist a year ago — that her COO didn’t put on the list because the JD doesn’t exist in the HRIS yet — is the flow designer. One hire. The most important on the plan. The reason the other eleven don’t need to be made.

Twelve becomes three. Headcount shrinks. Capacity goes up. Her COO will push back Monday morning, and he should — this is the hardest conversation a founder has with a great operator who built their career inside the old shape.

She’s having it anyway.

What to do now

1. Take your top five roles. List what each person actually does in a week. Not the JD — the calendar, the open tabs, the Slack channels, the recurring tasks. Be concrete: “reconciles the WIP every Friday morning.”

2. Map each item to a bucket. Deterministic, judgment-with-context, human-only. Force the call. First pass takes an hour per role. Second pass — where you correct the items you flinched on — takes twenty minutes.

3. Calculate the ratio. Any role more than 70% bucket 1 plus bucket 2 is not a role. It’s a flow with one human reviewer, and you’re paying full salary for it because nobody told you yet.

4. For your senior partners and client leads, ask the inverse question. Are they actually doing bucket 3, or buried in bucket 1 and 2 and only hitting bucket 3 in the gaps? Most senior people are buried in plumbing while expected to be the firm’s taste. The substrate’s point is to invert that ratio.

5. Now you have an org chart. Compare it to the one your COO sent you. The gap is your 2027 strategy. The next post — SOPs that run themselves — is how the flow designer actually builds the substrate that closes the gap. Read it before Monday.