{ "source": "Cooked Index — occupational AI risk register", "page": "https://cookedindex.com/jobs/door-to-door-sales-workers-news-and-street-vendors-and-related-workers/", "methodology": "https://cookedindex.com/methodology", "notice": "Verdicts are re-examined as evidence accumulates. Re-fetch before relying on this; the page above always carries the current score.", "scored_at": "2026-08-11", "model": "claude-opus-5", "occupation": { "title": "Door-to-Door Sales Workers, News and Street Vendors, and Related Workers", "soc_code": "41-9091", "category": "Sales", "us_employment": 2760, "median_annual_wage": 41380 }, "verdict": "EXPOSED", "risk_resistance": 43, "contested": false, "near_boundary": false, "dimensions": { "task_resistance": 13, "embodiment": 15, "liability_shield": 2, "trust_premium": 9, "judgment_accountability": 4 }, "reasoning": { "task_resistance": "A knock, a doorstep read, and a same-visit close cannot be delivered by software, but the parts around it — territory routing, lead qualification, pitch scripts, price quotes, and follow-up texts — are already handled by CRM and ad-targeting tools, which is why this sits at 13 rather than up with trades work that resists digitisation end to end.", "embodiment": "You are outdoors on foot in whatever weather, hauling product or pushing a cart, dealing with locked gates, dogs, stairs, and hostile residents on unfamiliar property — uncontrolled environments throughout — held just under the top band because the physical demand is walking and carrying rather than skilled manual technique.", "liability_shield": "There is no occupational license behind this work; the most you face is a municipal peddler or solicitation permit and a Do-Not-Knock registry check, and when a sale goes wrong it is the company's contract and the FTC three-day cooling-off rule that answer for it, not your signature.", "trust_premium": "Regular newsstand and street-cart customers, and route sellers who come back to the same blocks, build real repeat familiarity worth something — but most doorstep encounters are cold, one-shot, and with a stranger who did not ask you to come, so the relationship cannot be the product the way it is for an insurance agent with a book of clients.", "judgment_accountability": "Pricing, product line, financing terms, and refund policy come down from the employer or distributor; your calls are whether to keep talking, which house to try next, and how much stock to bring out — real choices, but low-stakes and reversible within a day, which is what puts this at 4." }, "rationale": "The core work — walking a route, knocking on doors, working a street cart, reading a stranger's face in the first three seconds and closing on the spot — is physical, in-person persuasion that language models cannot perform. But the shield is narrow: no license, no signature, minimal discretion, and the real threat to this occupation is not AI but e-commerce, targeted digital advertising, and app-based ordering, which have already shrunk US employment here to roughly 2,760. Lead lists, scripts, territory routing, and follow-up messaging are already generated by software.", "outlook": "The occupation keeps shrinking through 2035 — squeezed by online ordering, not by AI — with survivors clustered in high-ticket in-home sales and licensed product lines where a physical demo still closes the deal.", "what_would_raise_it": { "levers": [ { "dimension": "trust_premium", "change": "Growth of the segments where the in-person pitch IS the product — solar/roofing/pest-control canvassing, direct-to-farm and farmers-market vending, and utility/broadband 'switch' campaigns where regulators (e.g. state PUC door-to-door marketing rules) require an identifiable human agent to present the offer rather than a robocall or app funnel. Also visible: municipal street-vendor licensing regimes (NYC's 2021 permit expansion) that formalize and protect physical vending stalls, making the human seller a scarce, licensed presence.", "plausibility": "plausible", "would_add": 3 }, { "dimension": "liability_shield", "change": "State door-to-door solicitation statutes tightening so that the individual canvasser must be personally registered/permitted, badged, and named on the contract's right-of-rescission disclosure — with personal exposure for misrepresentation. Solar-financing fraud enforcement (California AB 1208-style proposals, FTC and multiple state AG actions against solar canvassers) is pushing toward per-canvasser registration rather than firm-level licensing.", "plausibility": "plausible", "would_add": 4 }, { "dimension": "judgment_accountability", "change": "If canvassers move from script-readers to on-the-spot qualifiers who must make and document the eligibility, affordability, or cooling-off disclosure call before a contract is signed — a duty some state solar and home-improvement rules already attach to the person at the door.", "plausibility": "plausible", "would_add": 3 }, { "dimension": "task_resistance", "change": "Task-mix shift: once list-building, routing, scripting, and follow-up messaging are fully automated, what remains is only the unscripted doorstep read and objection handling. This raises the share of the remaining day AI cannot do — but note it raises resistance per remaining worker while shrinking headcount, which is the pattern already underway here.", "plausibility": "already happening", "would_add": 3 } ], "ceiling_note": "The binding constraint is not AI capability but demand: e-commerce and targeted digital advertising have already reduced this to ~2,760 workers. Every lever above raises the per-worker score while the occupation continues to contract; a high resistance score on a vanishing job is not job security. No plausible route to a meaningful embodiment increase — the physical difficulty is already fully counted." }, "adjudication": { "method": "two independent runs agreed on the verdict", "outcome": "corroborated", "run_totals": [ 43, 48 ], "run_verdicts": [ "EXPOSED", "EXPOSED" ] }, "employment_history": { "points": [ { "y": 2017, "emp": 7520, "wage": 23810 }, { "y": 2018, "emp": 9430, "wage": 26430 }, { "y": 2019, "emp": 8930, "wage": 27420 }, { "y": 2020, "emp": 8360, "wage": 29730 }, { "y": 2021, "emp": 7860, "wage": 29390 }, { "y": 2022, "emp": 8640, "wage": 31100 }, { "y": 2023, "emp": 6220, "wage": 34910 }, { "y": 2024, "emp": 4590, "wage": 34530 }, { "y": 2025, "emp": 2760, "wage": 41380 } ], "from": 2017, "to": 2025, "change_pct": -63.3, "comparable_from": 2019, "spans_soc_revision": true }, "pivots": [], "license": "https://cookedindex.com/terms" }