E-commerce is facing a fundamental transformation – and most online shops have not yet noticed.
Jan Otto, Founder of JANO Consulting, describes in his guest article in the E-Commerce Magazine the transition to Agentic Search – an approach in which purchase intent, not keywords, drives search results. This approach becomes tangible in the shop as an Agentic Shopping Assistant and enables a form of AI Product Discovery, which replaces traditional filter categories and keyword searches. As a concrete real-world example, Jan Otto cites aginity.ai.
The End of the Keyword Era: Why Users No Longer Think Like Search Engines
Generative AI has fundamentally changed user behaviour. Customers today type natural language queries with context, budget and purpose – and expect the system to understand them.
Millennials prefer dialogue-based search
67 % der Millennials bevorzugen eine dialogbasierte, kontextuelle Suche gegenüber klassischer Stichwortsuche.
Higher Purchase Probability
Users who search in natural language have a 35% higher purchase probability.
Mobile search queries > 5 words
43 % aller E-Commerce-Suchanfragen auf mobilen Endgeräten enthalten heute mehr als 5 Wörter.
Why classical product search reaches its limits
Traditional search solutions are designed for keyword matches – not purchase intentions. In practice, this leads to three serious problems.
Zero-result problem
Over 70% of users leave the shop immediately when no results are displayed – even though the product would be available in the catalogue.
Relevance problem with complex queries
The more context a query contains – purpose of use, budget, body height – the worse classic search systems perform.
Return problem due to poor hit quality
Incorrect purchases caused by unsuitable search results generate returns. Each return costs 17–25 euros – regardless of the product price.
Generative AI as the New Infrastructure of Product Search
Agentic Search analyses purchase intent, extracts relevant product attributes and matches them directly against the product catalogue – structured, traceable and reasoned. A detailed technical comparison can be found in the article What is Agentic Search?.
Keyword Search vs. RAG vs. Agentic Search
A quick comparison of approaches for product search and consultation in online shops.
| Comparison Criterion | Keyword Search | RAG | Agentic Search |
|---|---|---|---|
| Data Basis | Static fields, keywords, filters | Top-k retrieval from text/vector index (subset) | Complete catalog match via structured product data points per SKU |
| Query Logic | Lexical matching (exact terms) | Similarity search | Attribute-based intent matching |
| Intent Understanding | Low for complex long-tail queries | Medium (depends on embeddings and prompting) | High (use case + attributes + follow-up questions) |
| Hallucination Risk | Low (minimal generation), but relevance gaps | Elevated (generative output) | Significantly minimized (structured product data) |
| Explainability | Medium (filters and scores visible) | Low to medium | High / attribute-based and traceable |
| Update Effort | Medium (synonyms/rules maintained manually) | High (re-indexing) | Low (live catalog) |
Agentic Commerce: When Every Product Becomes an Expert
In classic systems, the product catalogue is passive. In the agentic model, every product becomes active: it gets its own AI agent that independently responds to queries. Für Kunden wird dieser Ansatz als Agentic Shopping Assistant erlebbar.
Step 1 – Catalogue analysis
An agent generator analyses the entire product catalogue. For each SKU, all relevant attributes are captured, weighted and structured.
Step 2 – Intent extraction
An intent module analyses the query and extracts the purchase intent: What is the user looking for? Which attributes are important?
Step 3 – Agent activation
Based on the detected intent, only the product agents whose products are potentially relevant are activated.
Step 4 – AI Product Discovery with reasoning
Every search result comes with a justification at attribute level – transparent and traceable.
aginity.ai: The concrete example from the E-Commerce Magazine
Jan Otto explicitly names aginity.ai as an example of how Agentic Commerce already works in practice today. aginity.ai works with structured product data and attribute-based direct catalogue queries – traceable rather than purely probabilistic.
Minimised hallucination risk
Since recommendations are based on catalogue-based product attributes, the risk of misinformation is significantly lower than with purely generative RAG approaches.
Direct Catalogue Query at Attribute Level
aginity.ai works with structured product data and attribute-based direct catalogue queries – traceable rather than purely probabilistic.
GDPR-compliant in the EU
Hosted in the Frankfurt/Main data centre – designed for AI Act requirements.
Gemessene Ergebnisse aus laufenden Projekten
Strategic assessment: why now is the decisive moment
Companies that don't invest in intelligent product search now will fall behind – not in five years, but in two.
Customer expectations rise quickly
Kunden, die ChatGPT, Perplexity oder Google AI Overviews kennen, haben eine neue Benchmark für „gute Suche".
Competitors are catching up
Major e-commerce players are investing heavily in AI-powered search. For mid-sized retailers, Agentic Search is the most realistic way to keep up.
Integration window is now open
Systems like aginity.ai can be integrated into existing shop systems via an API. The effort today is 7 days.
Conclusion
Agentic Commerce is not science fiction – it is a proven technology that can go live in seven days: GDPR-compliant, with minimised hallucination risk through structured product data and measurable results from day one.
Frequently Asked Questions
What is Agentic Search?▾
Agentic Search analyses the purchase intent behind a query, matches relevant product attributes in the catalogue and delivers reasoned recommendations at attribute level. This improves transparency and minimises hallucination risks compared to purely generative RAG approaches.
What is AI Product Discovery?▾
AI Product Discovery describes the ability to identify the buyer's actual use case from a natural language query and find suitable products from the entire catalogue – intent-based, at attribute level, with explainable results.
What is the difference between Agentic Search and RAG?▾
RAG systems generate answers based on retrieved text passages from vector databases. Agentic Search works with structured product data and attribute-based matching – more explainable and with significantly minimised hallucination risk.
What is the difference between an AI shopping assistant and a chatbot?▾
A chatbot conducts open conversations. An Agentic Shopping Assistant delivers precise product results from the real catalogue – justified by attributes, no free conversation.
Do I need to switch shop systems for Agentic Search?▾
No. aginity.ai integrates as an API layer into existing systems. Shopify, Shopware and Magento are natively supported.
How long does integration take?▾
For a typical product catalogue with 1,000–50,000 SKUs, integration time is 7 business days – including setup, testing and go-live.
Are customer data stored or used for training?▾
No. Search queries are processed in pseudonymised form. No customer data leaves the EU data centre in Frankfurt/Main.
What does deployment cost?▾
aginity.ai offers three monthly packages: Agentic Chat (from 750/mo), Agentic Search (from 1,000/mo) and the Agentic Suite (from 1,500/mo) – each in the local currency plus VAT. All packages include 10,000 products and 10,000 queries/month.
Further content
About the author
Jan Otto is founder of JANO Consulting and advises mid-sized retail companies on digital transformation. His focus is on data-driven strategies for e-commerce, omnichannel and AI integration. This text was originally published as a guest contribution in E-Commerce Magazine (02.02.2026).