whirrwork
🌐 Traduction automatique. L'anglais est la source et peut être plus précis. Lire en anglais →

Robot ou humain ?

Warehouse order picking (AMR-assisted)

Picker walking with a cart vs Collaborative picking AMRs (people still pick)

Estimations modélisées par Whirrwork, pas des faits. Méthodologie · Bilan des prévisions

Humain · Par an338 587 $USManual cart picking
Robot · Par an371 332 $USAMRs as a service
Délai de retour sur investissementjamais
Année de bascule2029 (2028–après 2035)Base-case year (earliest to latest across scenarios)

On these numbers it is close: decide on the things that are not cost. Even running flat out, one robot would not displace enough work to pay for itself.

États-Unis · USD · BLS OEWS May 2025 national medians by SOC occupation

Faites vos propres calculs

Change anything. Results update as you type and stay in your browser; nothing is stored or sent. Works offline after your first visit.

Charge de travail
Hypothèse

Order lines picked per day.

Hypothèse

Operating hours per day (two shifts).

Hypothèse

Operating days per month.

Hypothèse

Active months per year.

Personnel
U.S. Bureau of Labor Statistics

BLS OEWS May 2025 national median hourly wage, SOC 53-7065.

U.S. Bureau of Labor Statistics

Wages and salaries were 70.0% of total employer compensation cost for private industry workers in June 2026 (ECEC), so total cost = wage / 0.70.

ilostat.ilo.org

Nominal 4.4% a year, 2015-2025 (ILOSTAT hourly earnings).

Hypothèse

Picks per hour for a person walking a cart. Varies hugely with order profile; use your own WMS figure.

Robot
Hypothèse

Picks per robot-hour. Collaborative AMRs do the walking; people still do the picking, so this is robot throughput, not autonomy.

CIVL

Reported Locus Robotics RaaS price per robot per month.

Hypothèse

WMS integration, mapping and change management for a first site. Integration often costs more than the robots.

Hypothèse

Contract or useful life in years.

Hypothèse

Share of the shift each robot is available.

Hypothèse

Share of robot work lost to faults.

Hypothèse

Share of picks that stay manual (oversize, hazmat, rush orders).

Hypothèse

Picker minutes per robot-hour. In assisted picking a person meets the robot at each pick face, so this is large by design.

Hypothèse

kWh per robot-hour.

Hypothèse

Electricity price per kWh (USD). Check your bill; energy is rarely a big line.

Finances
Hypothèse

Annual financing / discount rate. Nobody publishes yours; 8% is a round mid-2026 commercial borrowing guess.

Hypothèse

Years in the NPV comparison.

Hypothèse

Share of a human service price that is labour (rises with wages). Used to escalate and to adjust US service prices to other countries.

Hypothèse

Share of a robot service price that is labour (pilot, technician). The rest is equipment, which learning curves make cheaper.

Hypothèse

Share of a lease or RaaS fee that pays for hardware (learns down). The rest is service and support (moves with wages).

Download these results (CSV)

Résultats

OptionPar moisPar anPer unit (pick)Coût initialDélai de retour sur investissementBreak-even utilisation5-year NPV vs baseline
Humain Manual cart picking Référence 28 216 $US338 587 $US0,321 $US— — — —
Robot AMRs as a service
Robots needed: 5
 30 944 $US371 332 $US0,352 $US30 000 $US jamais not reachable
Utilisation 83 %
 -100 854 $US

What moves the answer most

Each bar shows the 5-year NPV of the best robot option when one input moves to the low or high end of its range, with the rest held at your values.

5-year NPV vs baseline: AMRs as a service5-year NPV vs baseline: AMRs as a serviceRobot output per hourlow value: 35 → -1 045 149 $UShigh value: 90 → 399 000 $USHuman output per hourlow value: 50 → 697 188 $UShigh value: 150 → -721 553 $USSupervisionlow value: 20 → 200 750 $UShigh value: 45 → -553 258 $USSubscription per robot per mo…low value: 1500 → 22 443 $UShigh value: 3000 → -347 446 $USHourly wagelow value: 14.36 → -185 906 $UShigh value: 21.54 → -15 801 $USWork lost to failureslow value: 0 → -73 709 $UShigh value: 0.08 → -182 286 $USUptimelow value: 0.8 → -100 854 $UShigh value: 0.97 → -2 217 $USIntegration and setuplow value: 10000 → -80 854 $UShigh value: 100000 → -170 854 $USEmployer on-cost multiplierlow value: 1.2790000000000001 → -145 492 $UShigh value: 1.579 → -56 215 $US-100 854 $USlow valuehigh valueSource: Whirrwork Robot-or-Human model v1.0.0 (whirrwork.com/tools/robot-vs-human/warehouse-picking/). Modelled estimate, not a fact.

Cost per unit to 2035

Real terms (today's money). Robot hardware follows Wright's law; wages follow each country's history. Shaded band: robots-slow to robots-fast scenarios. Estimations modélisées par Whirrwork, pas des faits.

ChartCoût humain par unité (Base), Coût robot par unité (Base), Coût robot par unité (Robots : progrès rapide), Coût robot par unité (Robots : progrès lent). Source: Whirrwork Robot-or-Human model v1.0.0 (whirrwork.com/tools/robot-vs-human/warehouse-picking/). Modelled estimate, not a fact.0,00 $US0,10 $US0,20 $US0,30 $US2026202720282029203020312032203320342035USD / pickscenario rangeCoût humain par unité (Base)Coût robot par unité (Base)Coût robot par unité (Robots : pr…Coût robot par unité (Robots : pr…Source: Whirrwork Robot-or-Human model v1.0.0 (whirrwork.com/tools/robot-vs-human/warehouse-picking/). Modelled estimat…

Année de bascule: 2029 (2028–après 2035)

What the spreadsheet misses

Sécurité
Less walking and pushing; mixed robot and forklift traffic needs rules.
Qualité
Fewer mis-picks are commonly reported because the robot screen shows the next pick; measure yours.
Flexibilité
RaaS lets you add robots for peak and hand them back, which is the main argument for it.
Edge cases
Oversize items, very low volumes and constantly changing slotting reduce the benefit.
Customer experience
Invisible to customers except via speed and accuracy.
Jobs and redeployment
Assisted picking keeps people picking; it changes the job (less walking) more than it removes it.
Réglementation
Workplace safety rules for mobile robots (ISO 3691-4 in many markets) apply to the vendor and the site.

Voir aussi

Sources de cette page

Les notes et les sources sont en anglais ; la version anglaise fait foi.

  1. Locus Robotics offers warehouse robots for $2,000 per month as a service CIVL, 2026-10
  2. Occupational Employment and Wage Statistics, May 2025 (median hourly wage by occupation) U.S. Bureau of Labor Statistics, 2025-05
  3. Employer Costs for Employee Compensation, June 2026: wages and salaries 70.0% of private-industry compensation (series CMU2020000000000P) U.S. Bureau of Labor Statistics, 2026-06
  4. Mean nominal hourly earnings of employees by sex and occupation (ISCO-08), DF_EAR_EHRA_SEX_OCU_CUR_NB ILOSTAT (International Labour Organization), from national labour force and earnings surveys, 2025
  5. Labour cost levels by NACE Rev. 2 activity (lc_lci_lev): Germany 2022-2025, total labour cost vs wages and salaries per hour Eurostat, 2025
  6. Rates and thresholds for employers 2025 to 2026 (employer Class 1 NIC 15% above £5,000) HM Revenue & Customs, 2025-04-06
  7. 令和6年賃金構造基本統計調査 (Basic Survey on Wage Structure 2024), table 1 by occupation Ministry of Health, Labour and Welfare, Japan (via e-Stat), 2024-06
  8. Average monthly earnings of employees by sex and economic activity, DF_EAR_EMTA_SEX_ECO_CUR_NB ILOSTAT (International Labour Organization), 2022
  9. Average annual wage of employed persons in urban units in 2024 (by position and by region) National Bureau of Statistics of China, 2025-05-20
  10. Euro foreign exchange reference rates (cross rates to USD), via Frankfurter European Central Bank, 2026-10-23
  11. Wright, T. P. (1936). Factors affecting the cost of airplanes. Journal of the Aeronautical Sciences, 3(4), 122-128 Journal of the Aeronautical Sciences, 1936
  12. Nagy, B., Farmer, J. D., Bui, Q. M., & Trancik, J. E. (2013). Statistical basis for predicting technological progress. PLoS ONE, 8(2), e52669 (62 technologies; Wright's law among the best hindcasters; errors grow with horizon) PLoS ONE, 2013