EIS-Webinar-AI-Search-Session-2-Product-and-ECommerce-Search-2024-10-09.pdf

LLM’s are becoming a more important part of an eCommerce strategy offering the ability to enhance product data attributes, improve data quality, provide recommendations and hyper personalization and enable conversational commerce.

In this webinar, we will explore how large language models (LLMs) are reshaping eCommerce strategies and the vital role they play in creating more dynamic, efficient, and personalized online shopping experiences.

How can LLM’s enable improved ecommerce search? This discussion will focus on how AI-driven models can deliver smarter search capabilities, offering users more accurate and personalized product recommendations while reducing the number of irrelevant results.

How can organizations use AI to remediate product data challenges?
In this segment, we’ll discuss how LLMs can identify and correct data errors, enhance product descriptions, and streamline data workflows to improve overall data integrity and relevance.

What is the best way to improve a contextualized and personalized experience?
This part of the discussion will dive into how to leverage LLMs to create contextual shopping experiences that adapt to the individual user’s needs, ensuring a more seamless and engaging journey from product discovery to purchase.

What is the role of a reference architecture in LLM-powered eCommerce?
This discussion will outline the importance of creating a well-structured architecture that ensures the seamless integration of AI models into business operations. We’ll explore key architectural components, how they interact, and why having a solid foundation is critical for long-term success in AI-powered eCommerce.

1. www.earley.com
WEBINAR WEBINAR
LLMs for Product and ECommerce Search
SETH EARLEY
CEO & FOUNDER
EARLEY INFORMATION SCIENCE
Media Sponsor
SANJAY MEHTA
SOLUTION ARCHITECT
EARLEY INFORMATION SCIENCE
PHIL RYAN
SVP STRATEGY & INNOVATION
LUCIDWORKS
PATRICK HOEFFEL
MANAGING PARTNER
PH PARTNERS

2. www.earley.com
Today’s Panel
[email protected]
https://www.linkedin.com/in/sethearley/
2
Seth Earley
Founder & CEO
Earley Information Science
Patrick Hoeffel
Managing Partner
PH Partners
[email protected]
https://www.linkedin.com/in/Patrick
-hoeffel/
Phil Ryan
SVP Strategy & Innovation
Lucidworks
[email protected]
https://www.linkedin.com/in/philryan999/
Sanjay Mehta
Principal Solution Architect
Earley Information Science
[email protected]
https://www.linkedin.com/in/sanjaymehta/

3. www.earley.com
Before We Get Started
WE ARE RECORDING
SESSION WILL BE
50 MINUTES PLUS
10 MINUTES FOR
Q&A
YOUR INPUT IS
VALUED
Link to recording & slides
will be sent by email after
the webinar
Use the Q&A box to
submit questions
Participate in the polls
during the webinar
Feedback survey afterward
(~1.5 minutes)
Thank you to our media partners : CMSWire and VKTR

3

4. www.earley.com
About Earley Information Science
4
Proven methodologies to organize information and data.
SELL MORE
PRODUCT
SERVICE
CUSTOMERS
EFFICIENTLY
INNOVATE
FASTER
1994
YEAR FOUNDED.
Boston
HEADQUARTERED.
50+
SPECIALISTS & GROWING.

5. www.earley.com
7 Part Search Series
5
Upcoming Sessions in this Series
Session 3: Oct 23 - Generative Engine Optimization (GEO): Revolutionizing SEO for the Future
Session 4: Nov 6 - Is my AI Assistant Lying to Me? Accuracy in Generative AI
Session 5: Nov 20 - The Practical Reality of AI and Large Language Models (LLMs) in
Transforming Business Operations
Session 6: Dec 4 - Vendor AI Strategies and Challenges: Lucidworks, Coveo, OpenSearch, and
Algolia
Session 7: Dec 18 - Stories of AI Impact on Real Peoples’ Lives and Livelihoods: The AI Gift
that Keeps on Giving
Session 1:Sept 25 – The Impact of AI on Search - Recorded
Session 2:Oct 9 - Product and Ecommerce Search

6. www.earley.com
Poll
6
1.Not on the radar
2.Planning stages for base AI powered eCommerce
3.Controlled experiments using AI in eCommerce
4.AI in eCommerce is currently banned
5.Implemented PoC’s (internal or externally facing)
6.AI in eCommerce applications deployed
7.None of the above
Where are you on your eCommerce AI journey?

7. www.earley.com
Abstract
7
Abstract:
LLM’s are becoming a more important part of an ecommerce strategy offering
the ability to enhance product data attributes, improve data quality, provide
recommendations and hyper personalization and enable conversational
commerce.
Key discussion points will include:
•How can LLM’s enable improved ecommerce search?
•How can organizations use AI to remediate product data challenges?
•What is the best way to improve a contextualized and personalized
experience?
•What is the role of a reference architecture in LLM powered ecommerce?

8. www.earley.com
Why are we here?
8
AI sector filled with hype and noise but potential is
very real.
Many vendors sell “aspirational functionality”
It is important to understand mechanisms and
limitations of Gen AI in eCommerce

9. www.earley.com
What can AI do for eCommerce?
9
Home Depot –
“Jake” vs
“Brandon”
Gen AI can give
you the
appropriate
experience

10. www.earley.com
AI Search/Retrieval Applications in eCommerce
10
•Acquisition – Advertising, SEO
•Product & Content Enrichment
•Search, Navigation, & Conversation
•Intent Detection & Personalization
•Recommendations & Suggestions
•Self Service, Troubleshooting
•Analytics & Insights

11. www.earley.com
Poll
11
1.Acquisition – Advertising, SEO
2.Product & Content Enrichment
3.Search, Navigation, & Conversation
4.Intent Detection & Personalization
5.Recommendations & Suggestions
6.Self Service, Troubleshooting
7.Analytics & Insights
8.Other
What applications are you considering for AI in
ecommerce?

12. Copyright © 2024Earley Information Science, Inc. All Rights Reserved.
What are the preconditions for an AI enhanced experience?
•Good product data and content are at the foundation
•AI can help with data enrichment
•Core mechanism is the ingestion of product data and
content into a space that can be queried – vector
embeddings enriched with metadata
•Other context – audience, standards, web pages, corporate
ontologies
12

13. www.earley.com
Data Remediation
13
With a reference architecture, Gen AI can be used to
develop product attributes and improve data quality

14. Copyright © 2024Earley Information Science, Inc. All Rights Reserved.14
Source Data to be Enriched
Original Product Name
Original Description
Attribute Names
Original
Catalog Context
SCHEMA/
MDM/PIM/ERP
Related Artifacts
Web Pages/Content
Digital Assets (images,
diagrams)
Existing Search Index
Product Sheets,
Manuals
Knowledge Base
Target Audiences /
Segments
Feed, API, Crawl, Direct
Merge by key – series,
category or attribute code
Unified Document Repository
Rules & Governance
Standards Bodies (ISO,
TC, SC)
Restrictions
Style Guide
Branding
Examples (Positive &
Negative)
Existing Search Rules
3rd Party Intelligence
Google Knowledge
API & Graph
Competitor Websites &
Search Results
Large Language
Models
EIS MRO Knowledge
Graph
User context
Signals / Telemetry
Search Analytics
Clickstream Analytics
Transactional
Analytics
Performance Metrics
KPI's
Audience / Segment /
Profile
Prepare/process:
normalize, classify, tag
Generate Doc
Embeddings
Create Graph Index
Generate Contextual Embeddings
(Application, Audience, Behavioral)
Generate Query / Phrase /
Q&A Embeddings
Generate Industry &
Organization Specific
Ontological Embeddings
Create Vector Index
Generate LLM Context and
Metadata

15. Copyright © 2024Earley Information Science, Inc. All Rights Reserved.
Sample Output Data
15
Category Code: "E3107000000“
Category Name: "Jumper Bars“
Brand Code: “ACM“
Brand Name: “ACME“
Series Code: 110400166560
Series Name Original: "Dedicated Short Bar“
New Name: “ACME Shorting Bar for Electrical Terminal Blocks“
New Description: "A dedicated shorting bar designed for use with
electrical terminal blocks to create a secure electrical connection and
prevent short circuits. Suitable for automotive, manufacturing, and
automation applications.“
Keywords: "shorting bar", "electrical terminal blocks", "short circuit
prevention", "automotive", "manufacturing", "automation“
Relevant Categories: "Electrical Components", "Terminal Blocks“
Reasoning: "The new name and description provide a clear
understanding of the product's purpose and applications, while the
keywords and categories ensure it can be easily found by the target
audience.”
Original New
•New names, descriptions and normalized attribute values will draw from both internal (ACME) sources
and publicly available data
•Output results will conform to ACME style guidelines and industry standards
•Enriched SEO information & metadata (e.g. reasoning) will be provided all output records

16. www.earley.com
Vectors can have thousands of dimensions
16
3 Dimensions: Location in physical space
Type: Restaurant
Cuisine: Italian
Rating: 5 star
Price: $$$
4 More Dimensions describe the entity
Your restaurant
You are here
How far apart are two vectors in
n-dimensional space?

17. www.earley.com
Embedding vs. Generative Models
17
Input Text
Embedding Model
-0.06352602690458298, -0.009375893510878086,
-0.07981908321380615, 0.010051253251731396,
-0.03623727336525917, 0.012990123592317104,
-0.06297729909420013, 0.03959296643733978,
-0.002809602301567793, -0.008157080039381981, ...
Text Embedding
Input Text
Generative Model Generated Text
"It was the
best of times"
“It was the best of times, it was the worst of
times…” is the famous opening line from
Charles Dickens’ A Tale of Two Cities.It
captures the dualities of the era and sets the
stage for the themes of revolution and
redemption in the story.
"It was the
best of times"
Size Options: 384, 512, 768, 1024, 1536, 2048, 4096 ...

18. www.earley.com
Embedding and Generative Models
18
Input Text
Generative Model Generated Output
"It was the best of times."
Suggest 5 options for titles of short stories
that begin with the quote above. The
audience should be young adults who are
struggling to find their way in the world.
For each title, include a 2-sentence
abstract that references the central
struggle of the protagonists, who are
grappling with various contemporary
topics involving relationships, education,
housing, inflation, political instability, social
unrest, changing technology, and climate
change.
“It was the best of times, it was the worst of
times…” is the famous opening line from
Charles Dickens’ A Tale of Two Cities.It
captures the dualities of the era and sets the
stage for the themes of revolution and
redemption in the story.
[ Context Window ]
[ Model Options ]
•GPT-3
•GPT-4
•GPT-o1
•Gemini-1.5
•Llama-3 (OS)
•Claude
•…
[ Response Options ]
•Length
•Style
•Tone
•Format
•Data Type

19. www.earley.com
How It Works
19
Product
Catalog
Embedding
Model
id: [-0.0362372735917, 0.01299592317104,
-0.0629772990013, 0.0395923733978, ...]
id: [-0.0362372735917, 0.01299592317104,
-0.0629772990013, 0.0395923733978, ...]
id: [-0.0362372735917, 0.01299592317104,
-0.0629772990013, 0.0395923733978, ...]
Vector
DB
User
Query
ID
ID
ID
ID
ID
Product
DB
Vector
DB
Embedding
Model
Generative
Model
ID
ID
ID
ID
Prod
Ans
Index Time:
Query Time:
Product Database

20. www.earley.com
Challenges of Gen AI in eCommerce
20
•Latency is higher than users typically expect in an eCommerce
experience
•Hallucinations & safety
•Cost & ROI
Solutions Available
•Offline processing of signals through LLM processor to pre-answer
common queries
•Conversational search, grounded in catalog data
•Dedicated Small Language Models (SLMs) to handle real-time queries
performantly.
•Chatbots and other modalities

21. www.earley.com
Opportunities of Gen AI in eCommerce
21
Paradigm Shift
•Better relevance, improved online buying experience
•Understanding user intent
•Opportunity to integrate non-product content and actions
into search results
•Generative marketing – customer acquisition
•Competitor analysis
•Greater levels of engagement

22. www.earley.com
Recommendations and Personalization
22
Using customers’ “digital body language” to inform
recommendations
Understanding prior interactions to contextualize
the experience

23. www.earley.comwww.earley.com
Copyright © 2022Earley Information Science, Inc. All Rights Reserved.Knowledge Graph
high-fidelity customer journey model
I RENEWI USE
customer model
What does it take to do this right?
I INSTALL I SHOP & I BUYI’M AWARE
23
INTELLIGENT
PERSONALIZATION
Component content model
User journey/customer model
Product data model
Knowledge architecture
Dynamic metadata identifies changing, real time. signals about
customer goals and intent while they go through their journey
Query: What kinds of
small appliances do you
have?
Action = Product
compare, purchase
Action = Download
installation guide
Action = Open offer email,
click through to site, click
offer
Dynamic customer model
Dynamic metadata:
campaign responses, click
through, recent purchases,
new goals change customer
metadata model, and
therefore audience
descriptors real time
Delivering Personalized Customer Experiences – At Scale
Tech docs
eCommerce
Support
LLM
CRM
Content
Product Catalog Ontology
Top of funnel content
(background on the issues
and challenges)
Product = Basic Widget
Offer = New customer
Middle of funnel content
(product selector,
comparisons)
Post purchase support
content (install guides,
troubleshooting info)
Post purchase nurture
content (how to get the most
from your Deluxe Widget)
Content type = White Paper
Topic = Predictive maintenance
Industry = Manufacturing
Stage = Awareness
Role = Technical
Content type = Product
compare tool
Topic = How to choose
Industry = Manufacturing
Stage = Shop
Role = Technical
Content type = Installation
guide
Product = Deluxe Widget
Content type = User tips
Product = Deluxe widget
Product = New and
Improved Super Widget

24. www.earley.com
Related Products
24
Compatible Products, Solutions, Kits, Bundles,
Accessories
Enable greater cross sell and upsell opportunities

25. www.earley.com
Help and Support
25
The customer journey frequently is a knowledge
journey.
Once we have great product data, other stages of the
customer journey can be enabled
Reduce the workload on sales and support by
accessing knowledge using RAG

26. www.earley.com
Next Steps
26
AI for eCommerce assessment
Identify use cases and opportunities for AI powered
interventions

27. Copyright © 2024Earley Information Science, Inc. All Rights Reserved.www.earley.comwww.earley.comwww.earley.com
27
Need Clarity on Retrieval Augmented Generation (RAG)?
https://www.earley.com/ama-article
Great companion to
The AI Powered
Enterprise
Download now:
or you can request a physical reprint

28. www.earley.com
Contact
[email protected]
https://www.linkedin.com/in/sethearley/
28
Seth Earley
Founder & CEO
Earley Information Science
Patrick Hoeffel
Managing Partner
Patrick Hoeffel Partners
[email protected]
https://www.linkedin.com/in/Patrick
-hoeffel/
Phil Ryan
SVP Strategy & Innovation
Lucidworks
[email protected]
https://www.linkedin.com/in/philryan999/
Sanjay Mehta
Principal Solution Architect
Earley Information Science
[email protected]
https://www.linkedin.com/in/sanjaymehta/

29. www.earley.com
29
We Make Information More Useable, Findable, And Valuable
Earley Information Science is a professional services firm headquartered in Boston and founded in 1994. With over
50+ specialists and growing, Earley focuses on architecting and organizing data – making it more findable, usable,
and valuable.
Our proven methodologies are designed to address product data, content assets, customer data, and corporate
knowledge bases. We deliver scalable solutions to the world’s leading brands, driving measurable business results.

30. www.earley.com
Vector Search
30
A vector is a mathematical representation of content and data.
Size
Style
Color
Medium, White, Polo shirt
Sm
Med
Lg
Polo shirt
Sweater
Dress shirt
Gray
Black
White

31. www.earley.com
Vector Search
31
A vector is a mathematical representation of content and data.
Latitude
Elevation
Longitude