Selling to the New Buyer: How AI Agents and Dialogue are Redefining eCommerce Search

At Balkan eCommerce Summit 2026, Paulius Nagys explored how AI-powered dialogue and product discovery are reshaping the future of eCommerce search. While acknowledging the growing influence of AI, the lecture also introduced a more cautious and realistic perspective on the current AI boom, comparing it to previous technology hype cycles. The session focused on the evolution of search behavior, the limitations of traditional eCommerce search engines, and how conversational AI and smarter discovery systems can improve customer experience and conversion rates.

AI Is Growing Fast – But the Industry Must Stay Realistic

Paulius opened the lecture with a skeptical but balanced view of the current AI landscape. He compared today’s AI enthusiasm to the dot-com era and the Java boom of the late 1990s, where companies received massive investments simply because they were associated with trending technologies.

According to the presentation:

many AI companies today still struggle with profitability,

despite enormous investments and attention.

He emphasized that:
LLMs and AI technologies are powerful tools,
but businesses should avoid treating them as:

“silver bullets” that will instantly transform everything.

The Real Problem in eCommerce Is Product Discovery

The lecture shifted toward one of the core challenges in eCommerce:

helping users find the right products quickly and accurately.

Paulius explained that:
search and product discovery remain unresolved problems despite more than 20 years of eCommerce evolution.

Traditional search systems still rely heavily on:

• keyword matching
• exact phrasing
• and structured product descriptions.

As a result:
many users fail to discover relevant products even when those products exist in the catalog.

Commerce Lost the Human Dialogue

One of the central ideas of the lecture was historical.

Paulius compared:
traditional physical stores,
where customers interacted with knowledgeable salespeople,
to modern supermarkets and online stores,
where customers are often left alone to navigate huge product catalogs themselves.

He explained that:

eCommerce removed the human conversation from shopping.

In physical stores:
sales assistants:

• recommended products
• asked questions
• understood uncertainty
• and guided purchase decisions.

Modern search systems, however:
expect users to already know:

exactly what they want.

Traditional Search Still Frustrates Customers

The lecture highlighted several major problems with standard eCommerce search:

• users type incorrect or incomplete keywords
• search engines fail to understand intent
• synonym recognition is weak
• discovery remains poor
• and customers abandon stores when they cannot find products quickly.

One example used:
a customer searching for “bed” instead of “sofa” on a furniture website may receive zero results because the search engine only matches exact keywords.

According to the presentation:

70% of eCommerce search tools fail on simple product queries.

The “Search Patience Gap”

One of the key concepts introduced was:

the “search patience gap.”

Using data collected from billions of search queries,
LupaSearch identified that:
many users repeatedly click multiple search results trying to determine whether they found the correct product.

Customers often:

• retry searches several times
• reformulate queries
• click multiple products
• and eventually leave the store entirely if unsuccessful.

The lecture explained that:

every failed search interaction increases the risk of losing the customer to a competitor.

Customers Expect Personalized Discovery

Paulius emphasized that:
modern consumers increasingly expect:

personalized shopping experiences.

Research mentioned in the lecture showed:

• over 70% of consumers expect personalization
• but most eCommerce stores still fail to deliver it effectively.

Despite widespread discussion around personalization,
the presentation argued that:

the actual implementation gap in most online stores remains enormous.

AI Can Improve Discovery – But Not Perfect Precision

The lecture distinguished between:

• AI-powered discovery
• and fully precise search results.

Paulius explained that:
LLMs are useful for:

• recommendations
• discovery
• conversational assistance
• and understanding intent,
  but they still suffer from:

hallucinations and imperfect accuracy.

Because of this:
AI should enhance product discovery,
rather than fully replace structured commerce logic.

Search Should Use Commercial Intelligence

A major insight from the lecture was that:
search engines should not only rely on product keywords.

Instead,
they should integrate:

• inventory levels
• profitability
• warehouse stock
• upcoming campaigns
• seasonal demand
• and commercial priorities.

For example:
if a retailer has high TV inventory before a football championship,
search results for “Samsung” should prioritize televisions rather than unrelated products.

This creates:

a balance between customer satisfaction and business profitability.

Dialogue-Based Commerce Is Returning

Paulius argued that:
AI-powered conversational commerce is effectively bringing back:

the shopping dialogue that disappeared decades ago.

Instead of forcing customers into rigid keyword searches,
future systems should:

• ask clarifying questions
• recommend products contextually
• understand uncertainty
• and behave more like human assistants.

The goal is not simply faster search,
but:

smarter product discovery that feels natural and supportive.

Businesses Often Underestimate Search Problems

The lecture also criticized how many organizations evaluate search performance.

According to Paulius:
IT teams often focus on:

• technical functionality
• autocomplete
• spell correction
• and system stability,
  while ignoring:

lost business opportunities caused by poor discovery.

Meanwhile:
design teams continue redesigning interfaces and running A/B tests,
while deeper search and discovery issues remain unresolved.

Key Takeaways

The lecture demonstrated that:
AI hype should be approached cautiously and strategically
LLMs are powerful tools but not magical solutions for every business problem
product discovery remains one of the biggest unresolved challenges in eCommerce
modern eCommerce removed the human dialogue from shopping experiences
traditional keyword-based search engines still frustrate many users
customers often abandon stores because they cannot find the right products quickly
the “search patience gap” highlights how users repeatedly retry searches before giving up
consumers increasingly expect personalized product discovery experiences
most eCommerce businesses still fail to deliver true personalization
AI is highly effective for discovery and conversational assistance, but not perfect precision
future commerce systems should combine AI with structured commercial logic
inventory, margins, seasonality, and business priorities should influence search results
conversational commerce is bringing back assistant-style shopping experiences
businesses often underestimate the revenue impact of poor search systems
the future of eCommerce search lies in dialogue, discovery, personalization, and commercially intelligent recommendations.