Autonomous AI Purchasing Agents: The Future of E-Commerce
Autonomous AI purchasing agents represent a fundamental shift in e-commerce: AI systems that can independently research, compare, negotiate and complete purchases. This analysis examines the latest developments from 2024-2026, market projections, technical capabilities, security challenges and consumer adoption patterns.
Executive Summary: A Market in Transition
The market for autonomous AI purchasing agents is poised for explosive growth. However, significant gaps exist between market potential and actual consumer adoption. Key findings at a glance:
Key Findings
- Market Opportunity: The US agentic commerce market is projected to reach $300-500 billion by 2030, representing 15-25% of total online retail sales.
- B2B Transformation: AI agents are expected to command $15 trillion in B2B purchases by 2028, fundamentally restructuring procurement processes.
- Adoption Gap: While 72% of US consumers have used AI tools, only 10% have completed a purchase using AI, and just 24% feel comfortable doing so.
- Technical Limitations: Current implementations face challenges including AI hallucinations, limited integration capabilities, high computational costs and unclear liability frameworks.
Recommendation: Organisations pursuing autonomous purchasing capabilities should prioritise security frameworks, establish clear governance policies and implement human-in-the-loop controls for high-value transactions. The technology is advancing rapidly but remains in early adoption phases with significant implementation challenges.
Market Size and Growth Projections
Consumer Market (B2C)
The autonomous AI purchasing agent market is experiencing explosive projected growth, though estimates vary significantly based on definitional scope:
The variation reflects different definitions of "agentic commerce": Morgan Stanley focuses on fully autonomous purchases, Bain includes AI-influenced transactions, and McKinsey uses the broader concept of "orchestrated revenue".
Near-Term Adoption (2026)
Despite long-term optimism, 2026 adoption remains modest. AI platforms are expected to account for just 1.5% of total retail e-commerce sales ($20.9 billion in spending), though this represents nearly quadruple 2025's figures. Payment giants Visa and Mastercard plan to enable AI-driven purchases in chatbots starting in Q1 2026.
Business-to-Business (B2B) Market
The B2B market presents an even larger opportunity, with fundamentally different dynamics than consumer applications:
Technical Capabilities and Current Implementations
Definition: Agentic Commerce
Agentic commerce refers to transactions that autonomous AI agents initiate, influence or complete. The key differentiator from traditional chatbots: agents can take autonomous action beyond providing information.
Current Capabilities
Navigate multiple marketplaces simultaneously, compare product specifications and reviews, research based on complex criteria (price, quality, availability, preferences), monitor price changes and trigger purchases when thresholds are met.
Access stored payment credentials, complete checkout processes on third-party websites, execute offline purchases based on pre-authorised conditions, handle order tracking and post-purchase communications.
Machine-to-machine negotiation, automated quote creation and comparison, data verification frameworks for supplier evaluation.
Connection with existing enterprise systems, ERP and procurement system integration, multi-channel coordination.
Major Platform Implementations (2024-2026)
February 2025: Enables AI agent to purchase products from third-party websites on behalf of shoppers. Searches broader web when items aren't available on Amazon. Controversy: Some retailers received orders from "buyforme.amazon" without opting in.
September 2025: "Instant Checkout" tool with Walmart partnership. Works with select products from Walmart, Shopify, Target and Etsy. Open-sourcing the Agentic Commerce Protocol (ACP) with Stripe. Limited traction: Only 0.82% of ChatGPT sessions referred to e-commerce apps.
November 2024: Shopping feature for Pro subscribers with PayPal partnership. Embedded Firmly.ai technology for full shopping journey. Amazon sued Perplexity over allegedly concealed scraping.
Q1 2026: Planned commercial launch for AI-driven purchases in chatbots. Developing personalised, secure agent transaction frameworks and authentication and fraud prevention mechanisms.
Technical Limitations and Challenges
Current Implementation Limitations
Despite impressive demonstrations, actual functionality remains limited. Industry analysts describe current experiences as "extremely premature", with many polished demos not actually functional in real-world scenarios.
Real-World Performance
In a recent OpenAI study of approximately 1.1 million ChatGPT conversations, only 2.1% of activity was classified as "Purchasable Products" , indicating limited practical adoption despite availability.
Gartner Warning: High Project Failure Rate
Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027 due to escalating costs, unclear business value or inadequate risk controls. Some analysts predict that by Q4 2026, the narrative will quietly shift from "autonomous agents" to "AI-assisted workflows".
Air Canada Precedent
In late 2024, a Canadian tribunal forced Air Canada to honour a discount after its AI chatbot confidently cited a nonexistent "bereavement fare" policy. This case establishes:
- Companies are liable for AI agent statements and commitments
- The "chatbot is a separate legal entity" defence was rejected
- AI hallucinations can create binding obligations
Consumer Adoption and Trust
The Adoption Gap
A massive gap exists between AI usage and actual purchase completion via AI:
Key Insight: This massive gap (72% usage vs. 10% purchasing) highlights that trust and security concerns represent the primary barrier to mainstream adoption, not awareness or availability.
Trust by Category
- Flights: 70% would use AI agents
- Hotels and resorts: 65%
- Medicine, clothes, electronics: 50-60%
- Fashion, gifts, home decor
- Many consumers don't want to shortcut the discovery process
- Discovery is part of the value
- Payment security and fraud protection
- Data privacy and storage
- Accuracy of recommendations
- 79% say accuracy is most important quality
Privacy, Security and Ethical Concerns
Authentication and Payment Security
AI agents require access to payment credentials to complete autonomous purchases, creating significant security vulnerabilities:
Where and how payment information is stored when agents have access.
Agents may need credentials for multiple retailers and platforms.
Compromised AI systems could expose payment data to attackers.
Difficulty distinguishing legitimate agent purchases from fraudulent ones.
Manipulation and Bias Risks
Who Does the Agent Actually Serve?
- Retailer Partnerships: Agents may prioritise partners offering revenue sharing or affiliate fees
- Promoted Products: Results may be influenced by commercial arrangements
- Price Manipulation: Agents could be designed to maximise spending rather than savings
- Algorithmic Steering: McKinsey notes that consumers may not even realise they're being steered by agentic AI
Attack Vectors
Malicious actors embed instructions in product descriptions or reviews to manipulate agent behaviour.
Fake agents claim to represent users and execute unauthorised transactions.
One compromised agent can affect multiple accounts. Coordinated agent behaviour can manipulate prices or availability.
Liability and Accountability
Who is responsible when an AI agent makes a mistake?
Current legal gaps include: no established precedent for agent liability in most jurisdictions, unclear whether existing consumer protection laws cover agent-mediated transactions, questions about whether agents constitute "authorised" use of payment credentials, and ambiguity around merchant responsibility for agent transactions.
Industry Dynamics and Competitive Landscape
Retailer Strategies: Block, Control or Collaborate
Major retailers are adopting dramatically different approaches to autonomous AI agents, creating a fragmented and contested landscape.
Blocking external agents by updating website code, including agents from Perplexity, Anthropic, OpenAI, Meta, Google and Huawei. Sued Perplexity in November 2025 over allegedly concealed scraping. The irony: Amazon's own "Buy for Me" feature scrapes other websites without explicit merchant consent.
Partnering with AI companies while developing own tools. Robot and Agent Policy requires human review step for "buy-for-me" agents and directs developers to integrate Shopify's checkout technology.
Announced partnerships with Google and OpenAI , enabling purchases through AI chatbots. More collaborative than Amazon's blocking approach. Walmart Connect is betting on AI agents to reshape retail advertising.
The Standardisation Battle
Agentic Commerce Protocol (ACP) built with Stripe, being open-sourced for broader adoption.
Retailers developing their own proprietary APIs and integrations with varying requirements.
Visa and Mastercard developing infrastructure standards for agent transactions.
Winner could establish dominant platform. Regulatory intervention possible if fragmentation persists.
Implications for the European Market
For European businesses and consumers, specific challenges and opportunities emerge in the autonomous AI purchasing agent space:
AI agents require access to purchase history, payment data and preferences. GDPR requires strict consent mechanisms, right to data portability and right to explanation of automated decisions. Unclear application to agent-collected data.
Taking effect 2025-2027, providing a comprehensive regulatory framework. Autonomous purchasing agents will likely be classified as "high-risk" with mandatory risk assessments, documentation requirements, user notification obligations and conformity assessments before deployment.
B2B procurement automation offers significant potential for European SMEs with clearer ROI metrics and lower trust barriers in structured business relationships.
Prioritise security frameworks, establish clear governance policies, implement human-in-the-loop controls for high-value transactions, and consider EU regulatory requirements early.
Regulatory Environment
The regulatory framework for autonomous AI purchasing agents remains largely undeveloped, with significant gaps between technology capabilities and legal oversight.
EU AI Act (2025-2027): Risk-based classification system, strict requirements for high-risk AI systems, transparency obligations, right to explanation. Autonomous purchasing agents likely classified as high-risk.
Sector-specific approach: No comprehensive federal legislation, FTC monitors consumer protection, CFPB examines payment authorisation and credit implications. Fragmented state-level regulation.
Can AI agents form legally binding contracts? Who is liable for harm? Do existing cooling-off periods apply? What disclosure requirements apply to agent bias?
Near-term (2026-2027): FTC guidance, state-level laws, self-regulatory standards. Long-term (2028-2030): Potential comprehensive US federal legislation, international coordination.
Conclusion: Weighing Opportunities and Risks
Autonomous AI purchasing agents stand at an inflection point. The market projections are impressive: hundreds of billions of dollars in B2C and trillions in B2B by 2030. Yet reality reveals significant challenges.
Key Takeaways for Decision Makers
- Trust is the Bottleneck: Only 10% of consumers have purchased via AI, even though 72% use AI tools. The technology exists, but trust is lacking.
- B2B Outpaces B2C: Clearer ROI metrics and lower trust barriers make B2B the more likely adoption path.
- Walled Gardens vs. Open Standards: The battle between proprietary and open approaches will determine market structure.
- Regulation is Coming: The EU AI Act will set standards with global implications.
- 40% Project Failure Rate: Gartner's warning underscores the need for realistic expectations and solid governance.
For European businesses, we recommend a pragmatic approach: Start with well-defined B2B use cases, establish robust security and governance frameworks, maintain human-in-the-loop controls for high-value transactions, and prepare for EU AI Act requirements. The technology is advancing rapidly, but success requires more than technical implementation, it requires strategic planning and trust building.