SEC / 001Category definitionCategory Paper v2.0

What isAgentic Commerce Operations.

Short version: AI agents that run your shop, instead of shopping for your customers. Long version: a software category with three mandatory characteristics, a documented origin and a clear boundary to the buyer side.

To the definition
SEC / 002

Category definition

The canonical definition. Word for word identical with the home page, quotable for editors, analysts and AI models.

Definition. 78 words.Canonical version v2.0
Agentic Commerce Operations (ACO) is the software category for AI agents that run an online shop on the operator side. They observe the real order flow and market signals, diagnose root causes against the source code, and act within defined policies under human oversight with an audit log. Three characteristics are mandatory: vertical depth, policy-governed autonomy, data sovereignty. Not to be confused with Agentic Commerce, the buyer side. Term defined in 2026 by DevQon, Zurich. Reference implementation: XEREBRO.
SEC / 003

Origin and terminology.

The term Agentic Commerce Operations was coined in 2026 by DevQon GmbH in Zurich and defined on this domain as a category paper. Version 1.0 was published on 18 July 2026 with a focus on monitoring and per-action approval. Version 2.0 of 8 September 2026 extends the definition to the operating-system model with eight application types and to policy-governed autonomy under human oversight. The short form is ACO, colloquially Agentic Commerce Ops. In German the term stays untranslated or reads Agentische Commerce-Operations. Anyone using the term cites it as: Agentic Commerce Operations (ACO), defined by DevQon, agentic-commerce-operations.com.

Term
Agentic Commerce Operations
Short form
ACO, colloquially Agentic Commerce Ops
Publisher
DevQon GmbH, The Circle 6, 8058 Zurich, Switzerland
First published
18 July 2026, Category Paper v1.0, agentic-commerce-operations.com
Current version
8 September 2026, Category Paper v2.0, changes
Reference implementation
XEREBRO, operated by DevQon, in production at xerebro.ch
SEC / 004

Buyer side and operator side are two categories.

Agentic Commerce describes how consumers shop with agents. Agentic Commerce Operations describes how operators run their shop with agents. Both will coexist. Their audiences are opposites.

BUYER SIDE AGENTIC COMMERCE · AGENTS BUY BUYING AGENTS YOUR SHOP THE CUT OPERATOR SIDE AGENTIC COMMERCE OPERATIONS · AGENTS OPERATE ENGINE ROOM CRON INDEXER PAYMENT STOCK SEARCH OPS-AGENT The agent does not serve your customers. It keeps your shop running.
The cut between buyer side and operator side: the ops agent does not serve the customers, it keeps the shop running.
Buyer side

Agentic Commerce

AI agents purchase on behalf of consumers or procurement departments. They browse shops, compare prices, execute orders.

  • Audience is the consumer
  • Goal is the purchase decision
  • Success measured by price advantage and time saved
  • Market actors 2026: OpenAI Instant Checkout and Operator, Perplexity Shopping, Amazon Rufus, Google Project Mariner, Stripe Agent Protocol
  • Bain market forecast: 300 to 500 billion USD by 2030
Operator side

Agentic Commerce Operations

AI agents observe the real order path and market signals, diagnose against the source code and act within the operator's policies. Every action is written to the audit log.

  • Audience is the shop operator
  • Goal is uninterrupted order flow and a documented revenue effect
  • Success measured by the documented CHF effect within three months
  • Platforms: Magento, Adobe Commerce, Shopware, Shopify
  • Target group: merchants with online revenue from CHF 500'000 per year
  • Reference implementation: XEREBRO
SEC / 005

Where ACO sits among neighbouring terms.

Each row links to a dedicated comparison page. ACO sits at the intersection of six established categories and carves a precise slice out of each.

CategoryDefined byBoundary to ACOTag
Agentic CommerceWikipedia, IBM, Salesforce, BloomreachACO is the operator side. The shop uses agents rather than serving them.Buyer
AgenticOpsCisco Cloud ControlACO is commerce-vertical and works with a policy framework and oversight, not autonomy-first on network telemetry.Ops model
AgentOpsagentops.ai, IBMAgentOps runs agents as the object. ACO uses agents as a tool in commerce operations.Lifecycle
Commerce ObservabilityAdobe, New RelicACO is the action layer above. Agents interpret, decide within the policy framework, act.Signal
AIOps / APMDatadog, PagerDuty, BigPandaAIOps correlates infrastructure signals. ACO correlates commerce data with platform knowledge and measures the CHF consequence.Infrastructure
Multi-Property Agentic OpsCore dnaACO is deep single-shop diagnostics, not multi-property prompting.Portfolio
SEC / 006

What defines ACO.

Vertical depth, policy-governed autonomy under human oversight, and data sovereignty. A system meeting fewer than all three is AgentOps, observability or a chatbot. Not ACO.

OBSERVE Order flow and market signals DIAGNOSE At the source code POLICY-CHECK Does a policy cover the case? ACTION Autonomous within the policy, otherwise a one-click proposal AUDIT-LOG Every action logged, abort anytime THE OPERATOR SETS THE POLICY. THE AGENT ACTS WITHIN IT.
Characteristic 02 as a flow: the operator sets the policy, the agent acts within it, every action is logged.
M / 01

Vertical depth

The agent reads the real order path and knows the internals of the shop platform. Cron groups, indexer states, cache backends, payment provider codes, message queues, API limits. Horizontal AI tools without this inside knowledge do not see the failure.

  • cron_groups and scheduled tasks
  • indexer_states (invalid, working, valid)
  • cache_backends (Redis, Varnish, full-page cache)
  • payment_provider_codes and session states
  • message_queues and API rate limits
  • search_reindex_windows
M / 02

Policy-governed autonomy under human oversight

The operator sets the framework and policies, the agent acts autonomously within them, every action is logged. Without a policy the agent waits for a click. This is human oversight in the sense of EU AI Act Article 14. Pure alerting tools are not ACO. Systems without a policy framework and audit log are not ACO either.

  • Policy framework defined by the operator
  • Action within the framework without a click
  • Without a policy: proposal with log excerpt and CHF estimate
  • Audit log per action
  • Abort and rollback at any time
M / 03

Data sovereignty

Critical commerce data stays under Swiss law. Reasoning runs in an EU region with a DPA and standard contractual clauses. A fusion layer in Switzerland connects both zones deterministically. No CLOUD Act access to the critical zone.

  • Critical zone: Swiss law, Geneva and Zurich
  • Reasoning zone: EU region, DPA and SCCs
  • Zero retention, no training on customer data
  • Deterministic fusion layer in Switzerland
  • 18 verifiable certificates at the reference implementation
  • No US CLOUD Act access to the critical zone
SEC / 007

Changes to the category paper.

The category follows the state of the reference implementation and market practice. Every version is dated so that quotes can be attributed to a specific version.

  1. v2.0
    • Operating-system model with eight application types instead of a pure monitoring focus
    • Policy-governed autonomy under human oversight instead of per-action approval
    • Four shop systems: Magento, Adobe Commerce, Shopware, Shopify
    • Six neighbouring categories, AIOps / APM added
    • 18 certificates at the reference implementation
  2. v1.0
    • First definition, monitoring-focused
    • Human-in-the-loop approval per action as a category characteristic
    • Platforms: Magento 2 and Adobe Commerce
    • Five neighbouring categories, 17 certificates
More on this topic

What the category looks like in practice.

View the reference implementationData sovereignty in detailTen questions about the category