Model Notes

← Back to calculator

What the model outputs. For a profession–country pair, the model returns $T_{50}$ — the median-like horizon (years) until AI/robots can perform roughly 50% of total task-time of that job in that country (at current adoption pace).

Global progress multipliers

KeyMeaningDefault
GAIrateGen‑AI progress multiplier (digital channel)1.00
GRPrateRobotics progress multiplier (physical channel)1.00

Parameter set (25 total)

Profession (16)

KeyMeaning
DigitDigitization of workload
StdTskTask standardization / repeatability
DataAccData availability / accessibility
DecmpDecomposability into micro-tasks
ErrTolError tolerance / detectability
RemotRemote-friendly work share
EcoStmEconomic incentive (wages/shortages)
PhysCtPhysical contact / manipulation
MobReqMobility / on-site logistics
EnvVarEnvironmental variability
LicnsLicensing / regulation
CapExCapital intensity for automation
DigShareShare of inherently digital tasks
HASHuman approval steps (1–5 scale)
TrustStakeholder trust in machine output
CustPrfCustomer preference for a human agent

Country (9)

KeyMeaning
CTAIdigDigital AI availability/adoption
CTAIphyRobotics capability/availability
LabEqRtLabor cost ratio
TechFrzCapital freeze / fleet stasis
OrgInrtOrganizational inertia
SocResSocial/cultural resistance
GovBrakPolitical/regulatory brakes
MktCmpMarket competition
MigOffsMigration offset (personal)

Computation pipeline

1) Base pressure

Two unit scores are computed from profession factors (weights sum to 1.00 in each block):

Digital amenability: $$ S = 0.22\,Digit + 0.18\,StdTsk + 0.12\,DataAcc + 0.12\,Decmp + 0.12\,ErrTol + 0.14\,Remot + 0.10\,EcoStm. $$

Physical/organizational friction: $$ F = 0.30\,PhysCt + 0.15\,MobReq + 0.12\,EnvVar + 0.18\,Licns + 0.25\,CapEx. $$

Base horizon (years): $$ T_{\text{base}} = T_0 \cdot e^{\,F - S}, \quad T_0=4.0. $$

2) Acceptance

Approvals and trust act multiplicatively (fewer approvals / higher trust → faster):

$$ \text{Agency} = \exp(-\gamma\,(HAS-3))\,\cdot\,\exp\!\left(-\lambda\,\Big(\tfrac{Trust+CustPrf}{2}-0.5\Big)\right), $$

with clamps to \([0.70,1.30]\), $\gamma=\ln(1.25)/2$, $\lambda=\ln(1.25)/0.5$.

3) Country frictions

Physical branch uses the full country brake: $$ \text{Common}_{phy} = \dfrac{\sqrt{SocRes\,\cdot\,GovBrak}}{MktCmp^{0.7}}. $$

Digital branch attenuates country frictions for highly digital jobs (DigShare close to 1):

$$ k_d = 0.5\,(1 - DigShare^2),\; k_m = 0.7\,(1 - DigShare^2), \quad \text{Common}_{dig} = \dfrac{(SocRes\,\cdot\,GovBrak)^{\,0.5\,k_d}}{MktCmp^{\,k_m}}. $$

4) Channels

Digital availability (clamped near 1 for cross-country parity), organizational drag and result:

$$ CTAI^{eff}_{dig} = (1-w)\cdot 1 + w\cdot \operatorname{clip}(CTAI_{dig};\,0.92,1.08),\; w=(1-DigShare)^{1.35}. $$

$$ \text{drag}_{dig} = OrgInrt^{\,1 - DigShare^{1.5}},\quad T_{dig} = \dfrac{T_{base}\cdot \text{drag}_{dig}\cdot \text{Common}_{dig}\cdot Agency}{GAIrate\cdot CTAI^{eff}_{dig}}. $$

Physical availability, economic freeze and result:

$$ CTAI^{eff}_{phy} = \min\!\left(1,\; \dfrac{CTAI_{phy}}{1+0.35\,\ln(LabEqRt)}\right),\quad Freeze = \dfrac{1}{\max(1-TechFrz,\;0.25)}. $$

$$ T_{phy} = \dfrac{T_{base}\cdot OrgInrt^{0.8}\cdot \text{Common}_{phy}\cdot Freeze \cdot Agency}{GRPrate\cdot \sqrt{CTAI^{eff}_{phy}}}. $$

If $TechFrz\ge0.80$, $LabEqRt\ge3$, $CTAI_{phy}\le0.15$: clamp $T_{phy}\in[20,35]$.

5) Aggregation & bounds

Weight the channels (emphasize digital as DigShare grows), apply migration offset and a global cap:

$$ w_{dig}=DigShare^{1.2},\quad T_{50} = \min\!\big(\,(w_{dig} T_{dig} + (1-w_{dig}) T_{phy}) \cdot MigOffs,\; 35\,\big). $$

Interpretation & calibration notes

Selected references

  1. Eloundou et al. (2023), GPTs are GPTs: labor market impact potential.
  2. ILO (2023), Generative AI and Jobs.
  3. IMF (2023), Labor‑market exposure to AI.
  4. Chen et al. (2024), Generative AI and Jobs.
  5. Zhang et al. (2025), AI Agents and work.