The org chart AI eats first: a role-by-role map.
The honest version of the AI-and-jobs story in ecommerce isn’t science fiction. It’s payroll math, and it’s already in the data. Stanford’s payroll-data study found employment for 22–25-year-olds in AI-exposed occupations down 13% since late 2022. Job postings for graphic artists fell 33% in 2025; writers, 28%. Meta’s automated Advantage+ campaigns passed a $20B run-rate. Intercom’s Fin resolves two-thirds of support tickets without a human. And the CEOs stopped being coy: Shopify’s Tobi Lutke now requires teams to prove AI can’t do a job before hiring for it. Here is the map of which ecommerce roles compress first, which transform, and which get more valuable — with the receipts.
- The compression is role-shaped, not industry-shaped: execution-layer roles (manual media buying, production copy, basic design, tier-1 support) are absorbing the hit first.
- Platforms did more than tools did: Advantage+ and Performance Max automated the median media buyer’s job from the supply side — 71% of Google advertisers run PMax; Meta wants “budget and a credit card” automation by end-2026.
- Support is furthest along: Fin at 67% average resolution; Salesforce cut 4,000 support roles; Klarna’s “700 agents” AI — and its partial walk-back — bracket the range.
- Counterweights are real: Yale finds no economy-wide disruption yet and warns of “AI-washing”; MIT found 95% of enterprise pilots deliver no P&L impact. The compression is concentrated, not universal.
- Net: the ecommerce org chart is collapsing from pyramids into pods — one senior owner per function, agents underneath.
What the data actually shows
Stack the evidence by layer. Macro: Stanford’s “Canaries in the Coal Mine” (ADP payroll microdata) shows the burden landing on the young — entry-level workers in AI-exposed roles down ~13%, while older workers in the same occupations grew; the ladder’s bottom rung is what’s breaking. Challenger tracked 54,836 AI-attributed US job cuts in 2025, then 87,714 in just the first five months of 2026 — 22% of all layoffs. Freelance (the leading indicator): writing gigs down ~33% since ChatGPT; company spend share on freelance marketplaces collapsed from 0.66% to 0.14% while AI-model spend rose past it. Ecommerce-specific: the platforms automated the buying layer themselves — Advantage+ at $20B+ run-rate growing 70% a year, PMax adopted by 71% of advertisers — while ~2M advertisers now use Meta’s generative creative tools. The media buyer didn’t lose to a startup; the auction absorbed the job.
The three layers of an ecom org, ranked by exposure
| Layer | Roles | 2026 status |
|---|---|---|
| Execution | Manual campaign ops, production copy, resizing/cutdown design, tier-1 CS, basic reporting | Compressing now — platform automation + agents; postings down 25–35% |
| Orchestration | Creative strategy, retention architecture, merch planning, CS escalation design | Transforming — fewer seats, each commanding agents; creative strategist demand rising |
| Judgment | P&L ownership, brand taste, offer architecture, negotiation, capital allocation | Appreciating — scarcer and better-paid as leverage per decision grows |
The middle row is where careers are decided this decade. A creative strategist who can brief, curate, and promote a hundred AI-generated variants a week is worth more than five production designers were in 2021 — and exactly that trade is happening: IAB found 83% of ad executives’ companies deployed AI in creative in 2025, up from 60% a year earlier, while demand for the strategist role climbs in every DTC hiring survey we can find.
AI didn’t eat the org chart top-down or bottom-up. It ate it middle-out from the execution layer — wherever work was definable enough to brief, it was definable enough to automate.
The honest counterweights
Two studies keep this piece from being a sermon. Yale’s Budget Lab finds no economy-wide AI labor disruption in the first 33 months post-ChatGPT — occupational mix is shifting within historical norms — and warns that companies “AI-wash” ordinary cost-cutting. MIT’s NANDA project found 95% of enterprise GenAI pilots produced no measurable P&L impact. Both are correct, and neither contradicts the role-level data: the disruption is real but concentrated — in young workers, execution roles, and the firms that actually operationalize the tools (vendor-bought deployments succeed ~67% of the time; internal builds ~22%). Klarna is the boundary marker: cut support headcount ~40% on AI confidence, then partially re-hired when satisfaction sagged. The technology compresses; the management decides how far without breaking the customer.
- We staff the judgment layer and rent almost nothing else: every LAMPWORK brand is a pod — senior owners with agents underneath (it’s LOS discipline #3).
- Acquisition math now includes an “automatable payroll” line: target SG&A that AI absorbs in 90 days is synergy we underwrite, conservatively.
- We hire people who already command the tools — the Klarna lesson is that taste and escalation judgment are the last mile, so we pay up for them.
Sources: Stanford via CNBC (Aug 2025); Bloomberry: 180M postings; Challenger, Gray & Christmas; Advantage+ run-rate; TechCrunch: Zuckerberg end-state; CNBC: Lutke memo; CNBC: Benioff; CNBC: Klarna; Yale Budget Lab; MIT NANDA via Fortune; IAB; Brookings: freelance evidence.
Own a brand this applies to?
We talk to founders at every stage — long before they're ready to sell.
Start a Conversation →