As the Founder and CEO of Dyver.AI, Octavian brings over two decades of entrepreneurial expertise in the software and eCommerce sectors to drive their mission forward. His journey from a curious software developer to an entrepreneur has given him a practical grasp of both technology and business, and he learned a lot about what makes the tech industry tick. He founded Dyver.AI with a simple belief: Product data shouldn’t slow eCommerce down. It should power growth. Manual product onboarding, data cleanup, and content enrichment have held eCommerce back for decades. Dyver.AI eliminates those bottlenecks through an AI-native platform that retrieves, normalizes, and enriches product data at scale. Their technology combines deterministic precision and generative intelligence to deliver clean, consistent, and SEO-optimized product content across every channel, automatically. Every processed listing makes Dyver.AI smarter, creating what they call a Living Product Data Environment – a continuously improving ecosystem that scales accuracy, visibility, and conversion for all users.
We Have an AI Spacecraft. Why Is Cross-Border E-Commerce Still Run on Spreadsheets?
Octavian Dumitrescu
Dyver.AI
Recap:
At Balkan eCommerce Summit 2026, Octavian Dumitrescu from Dyver.AI delivered a practical session on one of the biggest operational bottlenecks in cross-border eCommerce: product data.
The presentation explored why, even in the AI era, many online retailers still manage marketplace listings, product attributes, translations, descriptions, images, and category mapping through spreadsheets and manual work.
The key message was clear:
eCommerce is powered by product data, and AI can now automate much of the work that used to slow down cross-border growth.
eCommerce Runs on Product Data
Octavian began with a short history of commerce, from physical trade and catalog selling to online stores, marketplaces, mobile commerce, and modern omnichannel retail.
As commerce evolved, one thing became increasingly important: data.
Every product sold online needs structured information such as:
product attributes
titles
descriptions
images
videos
reviews
category placement
pricing information
SEO fields
translations
marketplace-specific requirements
In an offline store, customers can see and touch the product. Online, the product exists mainly through its data. If that data is incomplete, inconsistent, or poorly structured, the product becomes harder to find, harder to understand, and harder to buy.
Why Cross-Border Product Listing Is So Difficult
The main challenge is that every sales channel has its own rules.
An online retailer may need to sell through:
its own website
multiple marketplaces
local platforms
international marketplaces
price comparison sites
different country-specific channels
Each platform speaks a different “language” in terms of product categories, attributes, required fields, and formatting.
At the same time, suppliers also send data in different formats. Some provide structured files, others provide incomplete descriptions, poor images, or inconsistent product details.
This creates a complex operational problem: merchants must collect supplier data, clean it, enrich it, adapt it, translate it, and push it to multiple channels.
In many cases, this still happens manually through spreadsheets.
The Spreadsheet Problem
Octavian showed examples of marketplace product listing templates and explained how complex they can become.
Major marketplaces may have thousands of product categories and templates, including:
Amazon with around 25,000 categories
Allegro with around 23,000 categories
eMAG with around 4,000 categories
Scrooge with around 3,000 categories
Each category may require different attributes and specific product information.
For retailers with large catalogs, this means massive manual work, outsourced teams, repetitive data cleaning, and endless spreadsheet management.
The larger the catalog and the more markets a company enters, the harder the process becomes.
What AI Can Automate Today
A major part of the presentation focused on how AI can already support product data operations.
AI can help with:
automatic product categorization
extracting product information from images
tagging and sorting images
mapping attributes to marketplace requirements
generating structured product descriptions
creating meta descriptions and SEO fields
finding missing product information online
translating product content into multiple languages
generating lifestyle product images
creating product videos
placing products on models without photo shoots
changing product fabric or color visuals
These capabilities can significantly reduce manual work and help retailers scale product listings across more markets.
AI Must Follow Each Brand’s Workflow
Octavian emphasized that AI should not work as a generic one-size-fits-all tool.
Every merchant has different needs depending on:
catalog size
product categories
supplier data quality
marketplaces used
brand voice
industry standards
country requirements
internal processes
Sometimes a retailer already has product descriptions but needs translation. Sometimes they have only manufacturer attributes and need descriptions generated from scratch. Sometimes images need sorting, tagging, or enrichment before listings can be created.
This is why flexible AI workflows are critical. The system must adapt to each merchant’s actual use case.
Product Data Quality Matters More Than Ever
A key theme of the session was that product data quality is becoming even more important in the AI era.
In the past, merchants focused heavily on SEO. Today, product data must also support:
marketplace search
AI search
chatbots
recommendation engines
Generative Engine Optimization
digital sales assistants
Octavian explained that AI systems need accurate, structured product information in order to understand, recommend, and present products correctly.
If the data is poor, AI will not be able to represent the product properly.
This becomes especially important as chatbots and AI assistants become part of the shopping journey. Just like sales assistants in physical stores need product training, digital assistants need structured and accurate product data.
Generative Engine Optimization Requires Better Data
The presentation also connected product data to the future of visibility in AI-powered discovery.
As consumers increasingly use AI tools and generative search engines to find products, brands must ensure that their product data can be read, understood, and trusted by AI systems.
This means retailers need stronger product information, not just more product information.
AI discovery depends on:
clear descriptions
complete attributes
accurate categories
consistent images
structured metadata
reliable translations
updated availability and specifications
Without this foundation, products may become invisible in AI-driven shopping environments.
Trusting AI-Generated Data
Octavian also addressed a key concern: AI can make mistakes.
Because of that, AI-generated product data must be evaluated and validated.
Data quality depends on the specific merchant, platform, and marketplace requirements. A product listing that is “good” for one channel may be incomplete or incorrect for another.
To build trust in AI-generated data, businesses need evaluation systems that can check whether required information exists, whether attributes are correct, whether formats match marketplace rules, and whether the output aligns with brand and channel expectations.
The goal is not to remove human control completely, but to reduce repetitive manual work while maintaining quality.
Why This Matters for Cross-Border Growth
Cross-border eCommerce depends on the ability to list products quickly, accurately, and consistently across multiple countries and platforms.
If product data operations remain manual, expansion becomes slow and expensive.
AI can help merchants:
launch products faster
enter new marketplaces more easily
reduce manual listing work
improve product discoverability
localize content efficiently
maintain consistent brand communication
scale catalog operations without scaling teams at the same rate
For retailers with large catalogs, this can become a major competitive advantage.
Key Takeaways from the Session
The presentation delivered several important insights for eCommerce businesses:
eCommerce is fundamentally powered by product data
cross-border growth is often blocked by manual product listing processes
marketplaces have complex category and attribute requirements
spreadsheets are still one of the biggest operational bottlenecks
AI can automate categorization, attributes, descriptions, translations, images, and workflows
product data quality is essential for SEO, marketplaces, AI search, and chatbots
AI workflows must be adapted to each merchant, channel, and brand
human validation remains important, but repetitive work can be dramatically reduced
Perhaps the strongest message from the session was that businesses already have access to advanced AI capabilities, but many still operate cross-border commerce through outdated manual processes.
The opportunity now is to turn AI from a “spacecraft” idea into a practical operational engine for product data, localization, and scalable marketplace growth.