The leverage test.
Are you using AI for leverage, or is AI using you? The defining diagnostic of the Jobs Reshuffle Index. Two questions; both must clear.
Two questions, both must clear.
- New constraint. Does the worker own a new constraint that value has moved to? The three primitive constraints are risk (institutional / professional / legal liability), coordination (orchestration across teams, markets, or asynchronous components), and scarcity in what AI can't do (rare expertise, low-resource skill, judgment AI fails at).
- AI as ingredient. Is the worker bundling commoditized AI capabilities into their own offering - using AI to serve more customers, harder problems, or higher-trust outcomes than were tractable before? Or is the worker simply an input to someone else's AI?
Both legs must clear. Owning a new constraint without using AI is structurally durable but not above the algorithm - it's parallel to it. Using AI without owning a new constraint is being used by the algorithm - below.
The one-line check.
Use the directional shorthand:
- AI using you → below the algorithm. The worker is input to someone else's AI offering.
- You using AI → above the algorithm. The worker owns the offering; AI is rented infrastructure inside it.
The shorthand catches the most common confusion. A worker may use AI tools every day - the legal translator running CAT tooling, the radiologist with AI flagging - and still be below the algorithm if AI is the offering and the worker is the input layer. The arrow of dependency is what the test measures.
What "above" actually looks like.
The leverage test, applied to translator practitioners, surfaces three durable positions above the algorithm. Each is built around one of the three primitive constraints plus AI-as-ingredient:
- Risk owner - new accountability infrastructure. Builds accountability layers around AI-created risk. Example: Ariel Koren's Respond Crisis Translation - institutional accountability for MT deployment in asylum settings, an infrastructure that did not exist pre-AI.
- Customer + IP owner. Owns the offering itself - the customer relationship, the workflow, and the intellectual property that AI sits inside. Example: Gabriel Fairman's Bureau Works platform - a translator-founded software company that embeds AI capabilities under the founder's brand.
- Constraint-definition authority. Holds the social, professional, or institutional authority to ratify what counts as faithful, safe, or acceptable translation. Example: Corinne McKay and Damián Santilli's productized educator businesses - practitioner-to-practitioner authority over AI navigation in translation.
Translator scoreboard.
In the translator case study, the leverage test scoreboard reads 0 of 125 postings above the algorithm; 4 of 8 named practitioners above. The bifurcation is the central finding: industry job postings hire workers as inputs to someone else's AI; practitioners build their own offerings around new constraints. The pay differential - Human-in-the-loop floor at $5–$18/hour vs $40+/hour positional work - proves it empirically.