Technical Deep Dive
The core innovation of the agentic product discovery layer is its architecture, which is fundamentally different from traditional e-commerce search. Traditional search relies on inverted indexes, keyword relevance scoring (e.g., TF-IDF, BM25), and user click-through data to rank results. This system is optimized for human browsing: it returns a page of 10-20 results with images, titles, and prices, expecting the user to visually scan and select.
In contrast, an agentic discovery layer is built around intent parsing and parameter negotiation. The architecture typically involves three components:
1. Intent Parser: An LLM-based module that takes a natural language query from a shopping agent and converts it into a structured query object. For example, the query "Find me a lightweight laptop for coding under $1,200 with at least 16GB RAM and a good keyboard" becomes:
```json
{
"category": "laptops",
"price_max": 1200,
"ram_min": 16,
"use_case": "coding",
"features": ["lightweight", "good_keyboard"]
}
```
2. API Orchestrator: This module routes the structured query to multiple merchant APIs simultaneously. It handles authentication, rate limiting, and response normalization. It must translate the internal query format into each merchant's specific API schema—a significant engineering challenge given the lack of standardization across retailers.
3. Result Aggregator & Negotiation Engine: After receiving responses, this component de-duplicates, normalizes, and ranks results. Critically, it can engage in parameter negotiation: for example, if no laptop exactly matches the query, the engine might ask the merchant API for alternatives with slightly lower RAM or a higher price, or it might suggest a different brand. This is a form of automated haggling.
A key open-source project in this space is Agentic Commerce (github.com/agentic-commerce/agentic-commerce, ~4,200 stars). This repository provides a framework for building shopping agents that can interact with multiple retailer APIs. It includes pre-built connectors for major platforms like Shopify, WooCommerce, and Amazon (via their Product Advertising API), as well as a negotiation engine that uses a simple rule-based system for price and availability matching. Another notable project is ShopAgent (github.com/shopagent/shopagent, ~1,800 stars), which focuses on intent parsing using fine-tuned Llama 3 models and provides a benchmark dataset for evaluating agentic shopping performance.
| Component | Traditional Search | Agentic Discovery Layer |
|---|---|---|
| Query Interface | Keyword string | Structured JSON / Intent object |
| Data Source | Web index (crawled HTML) | Real-time API (JSON/GraphQL) |
| Ranking Signal | Click-through rate, PageRank | Intent match score, conversion probability |
| Result Format | HTML page with images | Structured product object (price, specs, availability) |
| Business Model | Pay-per-click (PPC) | Pay-per-conversion (PPCv) |
| Latency Tolerance | ~200ms (human patience) | ~500ms-2s (agent patience, can be batched) |
Data Takeaway: The shift from PPC to PPCv is the most consequential change. In traditional e-commerce, up to 98% of clicks never convert, meaning advertisers pay for massive amounts of wasted traffic. In an agentic model, the platform only gets paid when a sale happens, aligning incentives perfectly with merchants. This could reduce overall marketing costs by 30-50% for retailers, while increasing margins for the discovery layer.
Key Players & Case Studies
Several companies are already building components of this new layer, though none have fully captured the market yet.
Perplexity AI has been the most aggressive in pivoting toward agentic shopping. Their 'Buy with Pro' feature, launched in late 2024, allows their AI assistant to search for products, compare prices, and complete purchases using stored payment and shipping information. Perplexity uses a combination of their own web index and direct API integrations with retailers like Walmart and eBay. Their approach is notable for its transparency: the agent shows its reasoning steps (e.g., "Searching for 'waterproof jacket'... filtering by price... comparing reviews..."). This builds user trust, a critical factor for agentic commerce.
Shopify is positioning itself as the infrastructure provider for this new layer. In early 2025, they launched Shopify Agent Connect, a set of APIs specifically designed for AI agents. These APIs return structured product data, real-time inventory, and allow for direct checkout without a browser. Shopify's strategy is to become the default backend for agentic commerce, much like AWS became the default cloud for startups. They have also acquired a small startup called CartLogic (terms undisclosed) that specialized in intent parsing for e-commerce.
Amazon is a sleeping giant in this space. While they have not launched a dedicated agentic shopping product, their Amazon Product Advertising API is already one of the most comprehensive in the industry. Amazon's advantage is their massive logistics network and customer trust. If they launch an agentic shopping layer, they could immediately dominate by offering the lowest prices and fastest shipping. However, their current business model (heavily reliant on ad revenue from PPC) creates a conflict of interest—an agentic layer that optimizes for the best deal might cannibalize their ad business.
| Company | Product/Initiative | Key Strength | Key Weakness |
|---|---|---|---|
| Perplexity AI | 'Buy with Pro' | User trust, transparent reasoning | Limited merchant network |
| Shopify | Shopify Agent Connect | Merchant ecosystem, API-first | No consumer-facing agent |
| Amazon | Product Advertising API | Scale, logistics, trust | Conflict with ad business |
| Rabbit Inc. | Rabbit R1 (shopping mode) | Hardware + agent integration | Poor execution, limited adoption |
| Google | (Rumored) Project 'ShopBot' | Search data, AI expertise | Bureaucracy, privacy concerns |
Data Takeaway: The market is still fragmented, with no clear leader. The winner will likely be the company that can combine a large merchant network (like Shopify or Amazon) with a compelling consumer-facing agent (like Perplexity). The hardware players (Rabbit, Humane) have largely failed due to poor product execution, suggesting the battle will be won in software and APIs, not devices.
Industry Impact & Market Dynamics
The rise of agentic shopping will fundamentally reshape the e-commerce competitive landscape. The most immediate impact will be on digital advertising. Google's search ad business, which generated over $200 billion in 2024, is directly threatened. If AI agents bypass search results entirely and go directly to product APIs, Google loses its position as the primary gatekeeper of commerce. This is why Google is rumored to be working on Project 'ShopBot', an internal initiative to build an agentic shopping layer that integrates with Google Shopping and Merchant Center.
For small and medium businesses (SMBs), agentic commerce could be a double-edged sword. On one hand, it lowers the barrier to entry: a small boutique can expose its inventory via a simple API and be discovered by AI agents without needing to invest in SEO or paid search. On the other hand, it could lead to a 'race to the bottom' on price, as agents can instantly compare prices across thousands of merchants. SMBs will need to differentiate on unique products, quality, or service rather than price.
The market size for agentic commerce is projected to grow rapidly. According to industry estimates (compiled from multiple analyst reports), the total addressable market for AI-driven shopping could reach $150 billion by 2028, up from approximately $5 billion in 2024. This represents a compound annual growth rate (CAGR) of over 90%.
| Year | Agentic Commerce GMV (est.) | % of Total E-commerce |
|---|---|---|
| 2024 | $5B | 0.1% |
| 2025 | $18B | 0.4% |
| 2026 | $45B | 1.0% |
| 2027 | $90B | 2.0% |
| 2028 | $150B | 3.5% |
Data Takeaway: While the absolute numbers are still small relative to the $4.5 trillion global e-commerce market, the growth trajectory is exponential. If the technology matures and trust issues are resolved, agentic commerce could account for 10-15% of all online transactions by 2030, representing a $500B+ market.
Risks, Limitations & Open Questions
Despite the promise, significant challenges remain.
Trust and Security: The biggest barrier to adoption is consumer trust. Will users allow an AI agent to access their payment information, shipping address, and make purchasing decisions on their behalf? A single high-profile failure—such as an agent buying an expensive item without authorization or leaking payment data—could set the industry back years. Companies like Perplexity are addressing this by requiring explicit user confirmation for each purchase, but this defeats the purpose of full automation.
Merchant Adoption: For the agentic layer to work, merchants must expose high-quality, real-time APIs. Many smaller retailers still rely on outdated e-commerce platforms with no API access. While Shopify and WooCommerce make this easy, the long tail of independent merchants remains unconnected. This creates a 'chicken-and-egg' problem: agents are useless without merchant APIs, and merchants won't invest in APIs without a critical mass of agents.
Data Privacy and Vendor Lock-in: The agentic discovery layer will accumulate vast amounts of consumer preference data. This creates a new kind of data monopoly. If one platform (e.g., Perplexity or a future Google agent) becomes the dominant gateway, it could use this data to favor certain merchants, manipulate prices, or engage in anti-competitive behavior. Regulators are already circling this issue.
Technical Limitations: Current LLMs still struggle with complex multi-step reasoning and can hallucinate product specifications. An agent might confidently recommend a product that doesn't exist or misrepresent its features. The 'negotiation engine' is still primitive—it can handle simple price filters but cannot yet understand nuanced trade-offs like 'I'd rather have a better processor than more RAM.'
AINews Verdict & Predictions
Agentic commerce is not a fad; it is the inevitable next phase of e-commerce evolution. The shift from 'human-to-machine' to 'machine-to-machine' commerce is as fundamental as the shift from brick-and-mortar to online shopping was in the 1990s. However, the transition will be messy and contested.
Prediction 1: The 'Google of Agentic Commerce' will not be Google. Google's ad business creates an insurmountable conflict of interest. The winner will be a startup or a platform with no legacy ad revenue to protect. Perplexity has the best chance today, but they need to rapidly expand their merchant network and build trust.
Prediction 2: Shopify will become the dominant infrastructure provider. By offering agent-friendly APIs and a massive merchant base, Shopify is perfectly positioned to become the 'AWS of agentic commerce.' They will likely acquire a consumer-facing agent startup within the next 12 months to complete their stack.
Prediction 3: The pay-per-conversion model will become the standard. This aligns incentives perfectly and eliminates the massive waste of PPC advertising. Traditional ad networks will struggle to adapt, and we will see a wave of consolidation in the ad tech industry.
Prediction 4: Regulatory scrutiny will intensify. The collection of consumer preference data by a single agentic gateway will attract antitrust attention. We predict the EU will be the first to propose regulations requiring 'agentic neutrality'—meaning the discovery layer must not favor its own products or those of paying merchants.
What to watch next: The key metric to track is the number of merchants exposing agent-friendly APIs. If Shopify reports that 50% of its merchants have enabled Agent Connect within the next year, the tipping point is near. Also watch for the first major security incident—it will define the regulatory response for years to come.