Definition: Agentic Commerce
Agentic Commerce is the use of autonomous AI agents to independently initiate and complete transactions, such as payments and subscription renewals, for a person or organization.
Core characteristics of Agentic Commerce
Agentic Commerce turns a transaction into something an agent finishes end to end, not a form a human fills in.
- Autonomous execution: the agent selects, confirms, and pays without re-entering checkout details
- Bounded by a mandate: a signed instruction sets spending limits and approved merchants
- Machine-readable identity: a verifiable credential, not a shared login
- Cross-system reach: acts across a storefront, an ERP, and a payment network
Agentic Commerce vs. Conversational Commerce
Conversational commerce is a chatbot that recommends a product, then hands back to checkout. Agentic AI completes the purchase itself, within preset limits, so it is the agent who clicks “buy”.
Importance of Agentic Commerce in enterprise AI
Gartner projects AI agents will route over $15 trillion in B2B purchasing by 2028, 90% agent-assisted within three years. For Mittelstand firms this starts with an AI agent reordering consumables before stock runs low.
Methods and procedures for Agentic Commerce
Three building blocks make autonomous transactions possible.
Agent payment protocols
Network-level standards give agents a recognized way to pay: Visa’s Trusted Agent Protocol, Mastercard’s Agent Pay, Google’s AP2, and OpenAI/Stripe’s Agentic Commerce Protocol.
- Signed tokens replace a typed card number
- Network checks confirm the agent acts for a real account
- Fraud-liability rules are still forming
Mandate-based authorization
A mandate, signed once by a human, defines what an agent may buy, from whom, and up to what amount, checked on every transaction.
Agent identity and verification
Every transacting agent needs its own verifiable identity, per AI agent identity management, or a merchant cannot tell it from a compromised script.
Important KPIs for Agentic Commerce
Measuring Agentic Commerce means tracking autonomous spend, not just volume.
Operational metrics
- Autonomous transaction success rate: above 95%
- Mandate compliance rate: 100%
- Fraud and chargeback rate: at or below the human baseline
- Out-of-mandate escalations: reviewed weekly
Strategic metrics
McKinsey estimates agentic commerce could generate up to $1 trillion in US retail revenue by 2030, shifting spend toward agents.
Quality metrics
Quality means buying consistently from approved suppliers at negotiated terms; rising off-catalog purchases signal the mandate needs tightening.
Risk factors and controls for Agentic Commerce
Letting software move money raises the stakes of getting it wrong.
Fraud and unauthorized transactions
An agent that can pay becomes a target for manipulation.
- Prompt injection that leaks payment data
- Credential or token theft that impersonates the agent
- Runaway spending from a misconfigured mandate
Liability and legal attribution
Under German law, an AI agent cannot hold contractual liability, so AI liability and audit trails stay with the deploying company, especially since a manipulated agent can trigger GDPR liability.
Regulatory uncertainty
The EU has no agentic-commerce-specific rule yet. German consumer law (§ 312j BGB) requires an explicit “button solution” confirming payment, conflicting with a fully autonomous checkout, and human oversight duties under the EU AI Act still apply.
Practical example
A 140-employee precision parts manufacturer in Baden-Württemberg deployed an agent for reordering maintenance and repair supplies. A clerk once tracked stock across three warehouses by hand and often reordered late. The agent now monitors stock, compares two approved suppliers, and reorders within budget, escalating anything above it.
- Continuous stock monitoring across warehouses
- Price comparison against a supplier allowlist
- Automatic subscription renewal for consumables
- Monthly spend report with each transaction logged
Current developments and effects
The infrastructure for agentic commerce is moving faster than trust in it.
Protocol standardization
Card networks and AI labs are racing to set the default, with Visa, Mastercard, Google, and OpenAI each launching one since October 2025.
- Standards are converging on signed agent identities
- Networks are adding agent-specific fraud rules
- Checkout-ready integrations remain rarer in Germany than the US
Consumer trust gap
Only 9% of German consumers would let an AI agent handle a full purchase, per a 2026 ECC Köln survey.
Enterprise-side adoption outpaces consumer
Bitkom’s 2026 KI-Studie found active AI use in German firms rose from 17% to 41% in a year, with agents among the fastest-growing categories.
Conclusion
Agentic Commerce turns a transaction from something a person completes into something an agent completes under pre-set rules. Mandates and payment protocols are arriving faster than the legal and trust questions around them. For Mittelstand companies, the near-term opportunity sits in reordering and subscriptions, not open checkout. The mandate and audit trail matter more than transaction speed.
Frequently Asked Questions
What is the difference between agentic commerce and a shopping bot?
A shopping bot recommends a product, then hands back to checkout. Agentic Commerce completes the transaction, within an approved mandate.
Is fully autonomous AI checkout legal under German consumer law?
It is uncertain for consumer purchases, since § 312j BGB requires an explicit “button solution” confirming payment. Business purchasing under a pre-agreed mandate faces fewer constraints.
Does agentic commerce make sense for a company with around 100 to 200 employees?
Yes, usually for internal buying: reordering consumables, renewing subscriptions, or paying suppliers within a fixed budget.
How is GDPR compliance handled when an agent holds payment data?
The same obligations that apply to a human apply to the agent, with liability resting on the company, not the agent.
Who is liable if an AI agent makes an incorrect or fraudulent purchase?
The deploying company, since an AI agent cannot hold contractual responsibility under German law. A signed mandate and audit log matter more than its sophistication.
How does Superkind approach agentic commerce for Mittelstand companies?
Superkind builds AI agents connected to a company’s own ERP and CRM data that route purchase actions through existing approval rules, not open-ended autonomous spending.
