AI adoption has reached 88 percent, yet real delegation remains rare. This standard was written to close that gap.
"A digital employee is role-based, authorizable, closed-loop capable, evaluable, and governable. A chatbot, an avatar, an RPA script, or a copilot alone does not qualify."
Drawing on production deployments across Shopify merchants and verified industry sources, this standard defines how e-commerce digital employees are hired, authorized, evaluated, and governed.
The gap is delegation, not adoption
Organizational AI adoption reached 88% in 2025, yet AI agent deployment remains in the single digits across nearly all business functions (Stanford AI Index 2026). Most merchants have AI capability; few have governed digital labor.
Agentic commerce raises the stakes
McKinsey sizes agentic commerce at USD 3–5 trillion globally by 2030; Bain projects AI agents will influence or conduct 15–25% of US e-commerce sales. Merchants need operating infrastructure, not better chatbots.
Five conditions define a digital employee
Role-based, authorizable, closed-loop capable, evaluable, and governable. A chatbot, an avatar, an RPA script, or a copilot alone does not qualify.
Autonomy is granted per action, not per system
The L1–L5 levels attach to specific actions: the same customer-service employee may run logistics lookups at L4 while compensation commitments stay at L1/L3. Higher is not better.
Evidence comes before authorization
A five-stage evaluation chain — offline, sandbox, shadow run, gray release, production — determines what authority a digital employee may hold. A demo is not evidence.
Governance is the precondition, not the afterthought
Data minimization, least privilege, full audit trails, kill switches, rollback, and consumer protection are release prerequisites for L3 and above.
"Stronger capability requires stronger governance. Broader authorization requires stronger evidence."
From definition to deployment, the standard unfolds across eight layers.
Definition
What is a digital employee, and what is not. Terms, eight conformance criteria (DEF-001–008), and scope.
Role
Job descriptions for digital labor: objectives, SOPs, KPI/SLA, authorization boundaries, and accountable owners.
Architecture
A decoupled reference stack: intake, perception, knowledge, policy engine, decisioning, tools, audit, and human oversight. Models never write directly to production.
Capability
Six core capabilities with metrics and handoff conditions: perception, cognition, decision-making, execution, learning, and business outcomes.
Level
L1–L5 autonomy levels mapped to human oversight modes, upgrade conditions, and prohibited actions.
Evaluation
The five-stage evidence chain, plus runtime metrics across efficiency, quality, stability, safety, experience, and business value.
Governance
Seven control domains, P0–P3 incident classification, kill switches, and consumer protection.
Implementation
A 90-day blueprint, organizational roles, a maturity model, and continuous-improvement loops.
The six roles of an e-commerce digital workforce, each specified in full.
Each role ships with objectives, core responsibilities, key system interfaces, per-action autonomy levels, KPIs, handoff conditions, and required controls — the way you would write a job description for a human hire.
Where this standard comes from — and why you can trust it.
Verified industry sources
Findings are anchored to the Stanford AI Index 2026, McKinsey, Bain, and Visa, and aligned with NIST AI RMF and WEF agent-governance guidance. External figures are background evidence, never conformance thresholds.
Real deployments
Reference roles draw on production digital employees serving Shopify brands across apparel, jewelry, and consumer goods.
Open license
Published under CC BY 4.0: adopt, critique, extend, and translate freely with attribution.
Conformance classes C0–C4
From single components to production high-autonomy systems, so vendors can state exactly what they sell and merchants can state exactly what they buy.
