Task Tiers (T1/T2/T3) and Workforce Arbitrage
The question every agentic deployment must answer before any other is not ‘what can agents do?’ — it is ‘which tasks in this revenue loop are suitable for full agentic displacement, and what is the cost delta of achieving it?’ Task Tiers (T1/T2/T3) is the classification framework that answers the first part: every task in a business operation is T1 (fully automatable), T2 (augmented — requires human oversight), or T3 (human-mandatory). Workforce Arbitrage is the economic number that answers the second: the measurable cost delta between a human workforce and an equivalent agentic stack performing the same tasks. A market where 80 percent or more of the revenue loop is T1 produces a Workforce Arbitrage large enough to build a structural moat. A market below that threshold does not. The tier classification and the arbitrage calculation are not sequential steps in a process — they are the same evaluation, stated at two levels of abstraction.
What is the relationship between Task Tiers T1/T2/T3 and Workforce Arbitrage?
Task Tiers is Arco’s framework for classifying tasks by complexity and suitability for agentic deployment — T1 is fully automatable at a 37 to 50 times throughput advantage over human operators, T2 requires augmented human oversight, and T3 requires human judgement entirely. Workforce Arbitrage is the measurable cost delta between a human workforce and an equivalent agentic stack performing the same tasks — the economic number the tier classification produces. The tier composition of a market determines the magnitude of the Workforce Arbitrage: maximum at T1, partial at T2, near-zero at T3. Arco enters markets where T1 tasks represent 80 percent or more of the revenue loop — the threshold at which the Workforce Arbitrage is wide enough to sustain a structural moat.
When an operator approaches agentic deployment without the tier framework, the selection criterion defaults to visibility: which tasks are most manual, most time-consuming, or most complained about. These are not the same as the tasks where agentic deployment creates a structurally defensible economic advantage. A workflow that is high-volume and frustrating may still be T2 — it may require contextual interpretation, exception handling, or editorial judgement that keeps the human in the loop regardless of how capable the agent is. Applying agents to T2 tasks produces partial efficiency. It does not produce Workforce Arbitrage. The Task Tiers (T1/T2/T3) framework is the instrument that identifies which tasks in a Revenue Loop can be fully displaced — and which markets contain enough of those tasks to make the Breakable Market condition structurally sustainable. Without the classification, the agentic investment is real. The economic output is not.
The internal signature of a business that has deployed agentic capability without tier classification is a rising agent-to-task ratio alongside an unchanged human team. Agents handle the most automatable parts of a workflow. Humans handle the edges, the exceptions, and the decisions the agents surface for review. The Workforce Arbitrage never fully materialises because the human workforce is not displaced at T1 — it is repositioned at T2 and T3, where its cost is retained. The Labor-to-Compute Substitution that generates the full arbitrage requires full T1 displacement, not partial assistance. Partial assistance is a productivity improvement. It is not an economic restructuring. The Intervention Threshold distinguishes the two operationally: a system in which the steward intervenes because T1 tasks are escalating unnecessarily has a classification problem, not a deployment problem. The Operational Arbitrage Arco targets at the market level is only available when the tier composition of the revenue loop clears 80 percent T1 — the threshold at which Workforce Arbitrage is both large enough to matter and structurally defensible enough to hold against a competitor whose cost base is human.
The conventional agentic deployment pattern — insert agents into the most visible workflows, measure time saved, report adoption progress — consistently underdelivers because it does not apply the tier classification before deployment. A business that concentrates agents on T2 and T3 tasks because those tasks are high-value-looking is not reaching the economic ceiling the Stewardship Model defines for a T1-dominant operation. The Judgment Layer / Execution Layer distinction maps directly to the tier framework: T1 tasks belong entirely in the Execution Layer, where agents execute deterministically without human input at any step. T2 tasks straddle it — part Execution Layer, part Judgment Layer. T3 tasks are entirely in the Judgment Layer and should not be routed into the Execution Layer at all. A business whose agentic deployment is concentrated in T2 and T3 carries Administrative Density and Operational Drag from the coordination and oversight that T2 and T3 tasks require — without the economic offsetting effect of T1-level full displacement.
The tier evaluation is not a deployment checklist. It is a market entry filter. The Human-to-Logic Ratio is the upstream version of the same assessment: the higher the proportion of human judgement required in a market’s revenue loop, the lower the Workforce Arbitrage and the weaker the structural moat. When the tier classification confirms that 80 percent or more of the target market’s revenue loop is T1, the Architectural Certainty of the resulting build is effectively determined at market selection stage. The task composition guarantees that the system can reach the 72-hour MTTI threshold without T2 or T3 escalation as a structural feature of the design. The Revenue to Headcount Advantage — the 10:1 ratio Arco holds as a confirmed portfolio target — is the financial expression of this selection discipline applied consistently. It is not a target to optimise toward inside an existing build. It is the result of entering the right markets before the build begins. Memo #05 develops the full market selection framework: the conditions that make a market both T1-dominant enough to produce structural Workforce Arbitrage and defensible enough to sustain the moat once captured.
A business built into a T1-dominant market, with the Workforce Arbitrage calculated and confirmed before construction, has the conditions for Headcount Decoupling by design rather than by optimisation. Revenue scales with compute. Headcount does not scale with revenue. The 60 to 70 percent annual decline in LLM inference costs — confirmed in the Operational Arbitrage extended definition in Memo #05 — widens the Workforce Arbitrage over time without further investment. The business that captured the T1-dominant market first holds a moat that compounds annually on both sides: the agentic stack gets cheaper, and the human labour it displaced does not. Memo #01 develops the automated-to-autonomous distinction that the tier classification makes concrete: an Automated Business inserts agents into existing human workflows without changing the tier composition those workflows require. An Autonomous Business is built around a tier composition — 80 percent or more T1 — that makes full Workforce Arbitrage structurally achievable from day one. The classification precedes the construction. The construction realises what the classification made possible.
Technology changes what is possible. Task tier determines what it is worth.