The Nine Verbs — How to Break Down Work for AI Delegation

What This Means

Most AI deployment conversations fail before they start because no one can describe what work actually consists of. Teams talk about “automating processes” and “augmenting workers” without ever specifying what, precisely, the AI will do. The Nine Verbs solve this. They are the operational vocabulary of what Seampoint calls the Language of Work — a complete, non-overlapping set of operations that describe all work, every task, in every role, in every industry. Combined with the Four Platforms that define who performs work, the Nine Verbs give organizations a precise grammar for AI delegation. The question shifts from “Can AI do this job?” to “Which verbs does AI perform, and which do humans retain?” That shift changes everything.

The Precision Problem

Ask a hospital administrator which parts of radiology AI can handle and you will get a muddled answer about “supporting diagnostic workflows” or “enhancing clinician productivity.” Ask a bank executive where AI fits in lending and you will hear about “streamlining the loan process.” These statements sound reasonable. They are also useless. They describe nothing specific enough to design around, govern, or measure.

The problem is not a lack of ambition or intelligence. The problem is vocabulary. Organizations lack a shared language for describing what work consists of at the operational level. Without that language, every conversation about AI delegation floats at an altitude where nothing can be built, tested, or held accountable.

The Nine Verbs provide the missing vocabulary.

Nine Operations, No More

Work, at its most granular level, consists of exactly nine operations. Not eight, not twelve — nine. Each verb is defined by a clear test, and every test is mutually exclusive. A given task step is always one verb and never two.

Sense is the acquisition of raw data from the environment. A sensor reading, a camera capture, a blood draw, a form submission. The defining test: the output is raw data without interpretation. The moment meaning is attached, you have crossed into a different verb.

Interpret transforms data into meaning. A radiologist reading an X-ray, a financial analyst explaining a trend, an engineer assessing structural load from sensor output. The test: the output requires explanation or judgment. Interpret is where domain expertise lives.

Verify checks conformance to established criteria. Does this weld meet spec? Does this loan application satisfy underwriting guidelines? Is this patient’s lab result within normal range? The test: the output is pass or fail against known standards. Verify is not judgment — it is comparison.

Formulate creates new artifacts: plans, designs, strategies, compositions, prototypes. Writing a treatment plan, drafting a contract, designing a circuit board. The test: the output is something that did not exist before. Formulate is generative.

Decide selects a course of action and commits to it with accountability. Approving a loan. Authorizing surgery. Signing off on a product launch. The test is not cognitive complexity — it is liability. The actor who decides bears responsibility for the outcome. This distinction matters enormously.

Route directs work to its appropriate destination. Triaging an emergency room patient. Assigning a support ticket. Forwarding a referral to the right specialist. The test: the output is a destination or assignment, not a transformation of the work itself.

Transfer moves information between contexts without transforming it. Copying lab results into an electronic health record. Relaying a client request to a fulfillment team. Reading a purchase order aloud to a warehouse worker. The test: information is communicated or copied, not altered.

Actuate changes state in the physical or digital world. Administering medication. Pressing a button on a manufacturing line. Executing a trade. Writing a database record. The test: a state mutation occurs. Something in the world is different after actuation.

Monitor continuously observes for deviations from expected conditions. Watching vital signs. Tracking production line output. Scanning network traffic for anomalies. The test: sustained vigilance over time. Monitor is not a one-time check (that is Verify) — it is ongoing.

These nine verbs are exhaustive. Pick any task from any occupation in the Bureau of Labor Statistics’ O*NET database — all 18,898 of them — and every step in that task maps to exactly one verb.

A Sentence Structure for Work

The Nine Verbs gain their real power when combined with the concept of platforms — the actors that perform work, whether human, AI, or mechanical. Every step in any work process can be described as a sentence: Platform performs Verb on Object.

A human radiologist interprets a chest X-ray. An AI system senses transaction patterns. A loan officer decides on an application. A robotic arm actuates a weld.

This sentence structure — platform, verb, object — creates the atomic unit of work analysis. It is precise enough to design around, specific enough to govern, and structured enough to compare across roles, industries, and automation scenarios.

Walking Through a Real Workflow

Consider a commercial loan application moving through a regional bank.

A customer submits financial documents through an online portal. That is transfer — information moving between contexts. The bank’s document processing system extracts data from the financials. That is sense — acquiring structured data from unstructured inputs.

A credit model scores the application against underwriting criteria. That is verify — checking conformance to established thresholds. An analyst reviews the credit score alongside market conditions, the borrower’s industry trajectory, and qualitative factors that the model cannot capture. That is interpret — transforming data into meaning.

The analyst prepares a recommendation memo with proposed terms. That is formulate — creating a new artifact. The memo is forwarded to a senior credit officer. That is route — directing work to the appropriate decision-maker.

The credit officer reviews everything and approves the loan. That is decide — selecting a course of action with personal accountability attached. The approval triggers automated document generation and fund disbursement. That is actuate — changing state in the bank’s systems. Meanwhile, the bank’s risk management team tracks the loan’s performance against expected benchmarks. That is monitor — sustained observation for deviation.

Nine steps. Nine verbs. No gaps, no overlaps. And now the AI delegation question becomes precise: the bank can confidently hand off sense (document extraction), verify (credit scoring against criteria), transfer (document routing), and actuate (fund disbursement) to AI — liberating staff from coordination overhead. It should retain human ownership of interpret (contextual analysis), formulate (structuring the recommendation), decide (approval authority), and monitor (risk oversight that requires judgment about when deviations matter).

That is a different conversation than “AI will streamline our lending process.”

Why Decide Is the Critical Verb

Of the nine verbs, Decide occupies a unique structural position. It is the only verb where the defining test is not about the nature of the output but about who bears responsibility for it.

An AI system can rank options, calculate probabilities, and recommend courses of action — all of which fall under Interpret or Formulate. But the moment an organization needs someone to be accountable for the choice — someone who can be sued, fired, sanctioned, or imprisoned if the outcome goes wrong — that is a decision, and it requires a human.

This is not a capability gap that better technology will close. It is a structural constraint rooted in how legal systems, organizational governance, and social accountability actually work. An algorithm cannot be deposed in court. A neural network cannot lose its medical license. Until the legal and institutional infrastructure changes to accommodate non-human accountability, Decide remains a human verb.

Organizations that miss this distinction build systems where AI effectively makes decisions but no human is genuinely accountable for them — creating an accountability void. That is where regulatory, legal, and reputational risk concentrates.

The Judgment Question

A natural objection: “Where does judgment fit? Isn’t that the most important human capability of all?”

Judgment is not a tenth verb. It is a meta-capability that cuts across several verbs — primarily Interpret, Verify, Decide, and Formulate. When a physician interprets ambiguous symptoms, judgment is operating within the Interpret verb. When an underwriter verifies a borderline case that technically passes the criteria but feels wrong, judgment is operating within Verify. When an executive decides between two strategic options with incomplete information, judgment is operating within Decide.

This distinction matters because it explains why humans remain essential even as AI takes over more individual verbs. AI can verify faster and sense more data, but the application of judgment within those verbs — knowing when the standard criteria are insufficient, recognizing when context overrides the model — stays human. The Nine Verbs do not diminish the importance of judgment. They locate it precisely within the operations where it applies.

From Vocabulary to Strategy

The Nine Verbs are not an academic taxonomy. They are an operational tool. When an organization adopts this vocabulary, three things happen immediately.

First, AI deployment conversations become specific. Instead of debating whether AI can “handle” customer service, teams map the verbs in a customer service workflow and identify which ones are delegation-ready and which require human retention. The argument shifts from opinion to structure.

Second, governance becomes designable. Each verb carries different risk profiles and oversight requirements. Sense and Transfer are typically low-consequence and governance-safe. Decide and Interpret carry higher stakes. Mapping verbs to governance requirements produces a delegation framework that auditors, regulators, and boards can actually evaluate.

Third, workforce strategy gets honest. Instead of promising that “AI will free people to do higher-value work” — a claim that rarely survives contact with implementation — organizations can show exactly which verbs shift from human to AI, what volume of coordination overhead that represents, and what distilled human roles emerge around the verbs that remain. The result is role distillation, not job elimination: AI handles the handoff work, amplifies human judgment where it applies, and humans reserve the verbs where accountability and contextual reasoning are non-negotiable.

The Nine Verbs force precision. And in AI strategy, precision is the difference between transformation and expensive disappointment.

The Language of Work

The Nine Verbs are one component of a larger system. The Language of Work provides a complete architecture for describing and validating work allocation:

• Vocabulary: The Four Platforms define who performs work. The Nine Verbs (this page) define what operations work consists of.
• Grammar: The Capability Matrix defines which platform-verb assignments are structurally valid.
• Physics: The Physics of Work defines which assignments are sustainable given each platform’s architectural constraints.
• Compiler: The Compiler runs Grammar then Physics as a two-stage validation — catching delegation errors before deployment.

Related Concepts

• The AI Readiness Scale — The three categories of work that emerge from Language of Work analysis

Further Reading

• What Anthropic’s Deployment Data Reveals — How real-world AI deployment maps to specific operations, distinguishing core work from coordination overhead

Ready to apply the Nine Verbs to your organization’s workflows? Contact Seampoint for a work decomposition analysis that maps your highest-value AI delegation opportunities with structural precision.