Skip to main content
Back to home

Vocabulary

Plain-language terms for the AI workforce — Receipts, trust ladder, AI search visibility.

The words Crewmerce uses with you. One direct answer per term. Then the context that matters.

Hiring AI workers comes with vocabulary. AI workforce. Worker. AI agent. Receipt. Earned trust. Bounded spend. Work level. AI content labelling. Agentic commerce. Answer-first content. AI search visibility. Twenty-two terms across six groups — every one a plain-language answer first, technical alias second where useful, then the context that matters.

Product

Product

AI workforce

An AI workforce is a team of specialised AI workers — each one focused on a part of running an online store — that a small business hires to handle the repetitive jobs.

On Crewmerce, the AI workforce is the whole hireable roster. Order Manager. Customer Care. Product Specialist. Studio Photographer. Shipping Coordinator. Onboarding Specialist. API Integrator. SEO Specialist. GEO Specialist. Ads Specialist. Analytics Specialist. Business Advisor. The group of them, as a brand-level umbrella, is the AI Agents Crew. Each individual specialist is a Worker.

See also Worker · AI Agents Crew · AI agent

Product

AI agent

An AI agent is software that reads what is in your store, prepares an action, and waits for your approval — different from a chatbot, which only produces text.

Every Worker on Crewmerce is technically an AI agent. The agent reads from your shop, prepares an action under its current trust step, leaves a clear record of every action, and reports back. It is not a chatbot — a chatbot answers; an agent acts.

See also Worker · AI workforce · AI Agents Crew

Product

Worker

A Worker is one AI specialist you hire on Crewmerce — its own job, own opinion, own approval rules. The team of Workers is the AI Agents Crew.

When you talk to Order Manager, you are talking to a Worker. When you delegate "reword these listings," Product Specialist takes the job — that is also a Worker. The Worker is the indivisible unit of the AI workforce. The whole roster, as a brand-level umbrella phrase, is the AI Agents Crew.

See also AI agent · AI workforce · AI Agents Crew

Product

AI Agents Crew

The AI Agents Crew is Crewmerce's brand-level name for the team of AI workers you hire. Crew + Commerce = Crewmerce.

Each Crew member is a Worker — a self-contained AI agent with its own job, own opinion, and own approval rules. The Crew as a whole shares one chat surface, one Receipt format, one bounded-spend budget. "Crew" is the brand register; "Worker" is the specialist noun. Use Crew when you mean the team; use Worker when you mean one specialist.

See also Worker · AI workforce · AI agent

Product

Bounded spend

Bounded spend is the cost model that caps how much your AI workers can spend — per action, per day, per month — with a cost preview before any heavier work runs.

Simple tasks are included with your plan. Heavier work — bulk jobs, generation runs, multi-step builds — shows you a cost preview before it runs. You see the number, you approve, then it runs. Your monthly cap stops the total bill from being a surprise. See work level.

See also Work level · Receipt · Earned trust

Product

Work level

Technical alias: Token tier (T1 · T2 · T3)

Work level is the three-step cost classification Crewmerce applies to every worker job: Read (free), Draft & prepare (included with your plan), Generate & automate (heavier work — shown as a cost preview before it runs).

Read covers status checks, audits, lookups — anything where the worker just reads from your store. Draft & prepare covers drafted replies, proposed rewrites, single-product synthesis. Generate & automate covers bulk content, photo generation, multi-step campaign builds and other heavier jobs. Each worker capability is tagged. See bounded spend.

See also Bounded spend · Receipt

Autonomy

Autonomy

Earned trust

Technical alias: Earned autonomy

Earned trust is Crewmerce's four-step safety model. Every worker starts by reading and preparing — never changing anything. You approve every action that matters. Workers earn the ability to handle repeated low-risk jobs on their own only through actions you have already approved.

Workers do not arrive autonomous. They start in read-only mode or suggested-action mode, prepare every action with the full proposal and the rule that triggered it, and leave a clear record after each approval. Approved automation is earned per capability — not per worker — through actions you have already said yes to. Refunds, bulk changes and anything sensitive stay approval-first. Trust compounds with evidence.

See also Read-only mode · Suggested action · Approved automation (light) · Approved automation (broader)

Autonomy

Read-only mode

Technical alias: Ring 0 · Observe

Read-only mode is the starting position for every worker. The worker reads from your shop and proposes — it never changes anything.

Analytics Specialist and Business Advisor default to read-only forever — their job is to read and advise, never to act. Every other worker can be paused in read-only mode if you want it observing without proposing yet. See earned trust for the full step lifecycle.

See also Earned trust · Suggested action · Approved automation (light)

Autonomy

Suggested action

Technical alias: Ring 1 · Suggest

Suggested action mode means the worker prepares each action — the proposed reply, refund, rewrite or shipping step — and waits for your tap before anything ships.

Most execution Workers — Order Manager, Customer Care, Product Specialist, Studio Photographer, Shipping Coordinator — default to suggested-action mode. You see what is about to happen. You approve. The action ships with a clear record. Nothing runs behind your back. See earned trust.

See also Earned trust · Read-only mode · Approved automation (light)

Autonomy

Approved automation (light)

Technical alias: Ring 2 · Act-and-Report

Light approved automation lets a worker run low-risk repeats of actions you have already approved many times — under your rules, with every action still showing up in your receipt log.

Light automation is earned per capability through enough approved suggested-action runs of the same kind. You grant it explicitly — never the whole worker, only the capability that proved itself. The worker executes first and shows the Receipt immediately; you can revoke any time. See earned trust.

See also Earned trust · Suggested action · Approved automation (broader)

Autonomy

Approved automation (broader)

Technical alias: Ring 3 · Autonomous

Broader approved automation is a ceiling for advanced cases — never the default, never granted by Crewmerce, never applied to bulk or financial actions.

Broader automation means the worker plans and acts without a per-action approval, still leaving a clear record and respecting bounded spend. We do not advertise it as a milestone. Some capabilities will never become eligible (refunds, bulk pricing changes, financial transfers). The floor stays safe; the ceiling stays reserved. See earned trust.

See also Earned trust · Approved automation (light) · Read-only mode

Provenance

Provenance

Receipt

A Receipt is the clear record left after every worker action — what was found, what changed, why and when. Retained forever.

Every refund posted, every label printed, every description rewritten, every image generated — each one ships with a Receipt. The Receipt makes the action auditable; the action record makes the whole history searchable. Receipts are the metaphor moat — black-box AI does not give you them; Crewmerce ships proof.

See also Action record · Earned trust · AI content labelling

Provenance

Action record

Technical alias: Evidence Trail

The action record is the searchable log of every receipt your AI workers have ever produced — by worker, by job, by time, by cost.

When a worker acts, the action is committed to the receipt log. The action record is that log surfaced as a timeline, a per-worker activity feed, an audit view. Use it to answer "who refunded that order," "why did the worker do that," "how much did this campaign cost."

See also Receipt · Earned trust · Bounded spend

Provenance

AI content labelling

Technical alias: Provenance

AI content labelling means every AI-generated image and rewrite is labelled and trackable, so downstream tools can verify the AI origin of the content.

Every AI-generated artifact on Crewmerce ships with content labelling default-on. Images carry an image provenance manifest plus an AI image watermark. Text turns carry a first-exposure disclosure surfaced before publish. Labelling is one of the load-bearing requirements of the upcoming EU AI transparency expectations. Receipts are the per-action proof; labelling is the per-artifact proof.

See also Image provenance manifest · AI image watermark · EU AI labelling rules

Provenance

Image provenance manifest

Technical alias: C2PA · Coalition for Content Provenance and Authenticity

Image provenance manifest is the machine-readable signature embedded inside every AI-generated image — it says what produced the image, when, on whose behalf.

Every image Studio Photographer generates embeds a provenance manifest naming the AI capability that produced it, the generation timestamp, the input reference, and the Crewmerce signing identity. Downstream tools that read the manifest can verify the asset was worker-generated. The manifest is one of the two labelling signals on every AI image.

See also AI content labelling · AI image watermark · EU AI labelling rules

Provenance

AI image watermark

Technical alias: SynthID

AI image watermark is a perceptually invisible mark embedded in the pixel domain of every AI-generated image — robust to compression, cropping, and most edits.

Every image Studio Photographer generates carries an invisible watermark alongside the image provenance manifest. Together they give two independent labelling signals — one cryptographic (the manifest), one pixel-domain (the watermark) — so a downstream consumer can verify the image's AI origin even if one signal is stripped.

See also AI content labelling · Image provenance manifest · EU AI labelling rules

Search & AI search

Agentic commerce

Agentic commerce is the emerging category where AI agents — not browsers, not humans — discover, evaluate, and transact with merchants on behalf of end users.

When a shopper asks an AI assistant to "find me a bracelet that matches my outfit," the model becomes the buyer's agent and the merchant has to be discoverable in AI search, readable in machine formats, and transactable through agent-facing payment rails. Crewmerce ships the AI workforce that lets a shop participate.

See also Answer-first content · AI search visibility · One-call integration pattern · AI client integration

Search & AI search

Answer-first content

Technical alias: AEO · Answer Engine Optimization

Answer-first content is content structured so AI assistants (ChatGPT, Claude, Perplexity, Google AI Overviews, Bing Copilot) cite it as a direct answer to user questions.

Where traditional SEO optimises for a ranking inside a list of blue links, answer-first content optimises for citation inside a synthesised answer. Each answer leads with the direct answer in the first 1-2 sentences, then provides context. GEO Specialist's job is answer-first content plus entity density per page, schema graph integrity, and llms.txt hygiene.

See also AI search visibility · Agentic commerce

Search & AI search

AI search visibility

Technical alias: GEO · Generative Engine Optimization

AI search visibility is the discipline of making a site visible inside generative AI search outputs — through structured data, entity density, llms.txt and AI-citation tracking across ChatGPT, Claude, Perplexity, and Google AI Overviews.

AI search visibility encompasses answer-first content (the content discipline) plus the machine-readable substrate that lets AI crawlers cite a brand confidently. Crewmerce ships GEO Specialist as a hireable Worker — the operation that built Crewmerce's own AI-search surface (this page, llms.txt, the schema graph at /how-it-works) is the same operation the Worker runs for customer shops.

See also Answer-first content · Agentic commerce

Architecture

Architecture

AI client integration

Technical alias: MCP · Model Context Protocol

AI client integration is the standard for letting an AI assistant connect to an external system's tools and data — a typed bridge between an AI client and a service's capabilities.

Crewmerce ships an AI client integration server so any compatible client (ChatGPT Desktop, Claude Desktop, Perplexity and similar) can hire and operate the AI workforce inside that client's chat. The integration is the transport; the worker is the same worker.

See also One-call integration pattern · Agentic commerce

Architecture

One-call integration pattern

Technical alias: UCP · Universal Connector Pattern

One-call integration pattern is the Crewmerce design rule that says every external surface — dashboard, AI client, Slack, browser extension, future connectors — wraps the same single call into the workforce.

The pattern guarantees one entry point per worker per connector. There is no worker-by-connector wiring matrix. Add a new connector? It calls the single workforce entry point. Add a new worker? Every connector picks it up automatically. This is what makes the workforce extensible without exponential integration cost.

See also AI client integration · Agentic commerce

Compliance

Compliance

EU AI labelling rules

Technical alias: EU AI Act (effective 2 August 2026)

EU AI labelling rules are the EU's upcoming transparency requirements for AI-generated content — machine-readable labelling, clear human-facing disclosure, transparent logging — designed to take effect mid-2026.

Crewmerce is built with these expectations in mind. Images: provenance manifest plus watermark. Text: first-exposure disclosure surfaced before publish. Decisions: a clear Receipt on every action, action record as the transparent log. Training data: zero use of customer data — past, present, or future.

See also AI content labelling · Image provenance manifest · AI image watermark

Behind the scenes

Crewmerce manages a managed AI execution layer behind every worker — the right AI capability is selected per task. We pick the providers so you do not have to. See /how-it-works for how that selection works under the hood.

Ready when you are

Hire your first AI worker.

Free tier is read-only across the workforce — zero cost, zero risk. Grant workers more freedom once you trust them.

Vocabulary — AI workforce terms, the trust ladder, Receipts and more, defined — Crewmerce