Transforming Content Operations in the Age of AI

Transforming Content
Operations in the Age of AI
Rebecca Wyatt, Enterprise Knowledge
Elliott Risch, Enterprise Knowledge
KMWorld, Nov 2025
Washington, DC

Presenters
PARTNER AND DIVISION
DIRECTOR,
ENTERPRISE KNOWLEDGE
REBECCA
WYATT
TECHNICAL CONSULTANT
SEMANTIC & AI ENGINEERING,
ELLIOTT
RISCH

Today’s Agenda
Transforming Content Operations in the Age of AI
Enterprise Knowledge
OVERVIEW &
APPROACH
DEFINING GOOD
AI OUTCOMES Q&A
WHAT MAKES
CONTENT AI
READY

DRC Product Value Statement
The Intel Digital Resource Center (DRC) is the learning and resource portal to
support Intel's digital content operations executed by employees, contractors,
and external agency partners.
The DRC enables learners to efﬁciently ﬁnd, implement, and govern a
dynamic set of digital services and tools, while exercising best practices.

Outcomes
This project allowed Intel to:
⬢ Align around technical recommendations to take the
POC to production;
⬢ Learn how semantic search can be leveraged in the
future on Intel.com; and
⬢ Prioritize content and semantic remediation activities
which will help Intel to prepare content for AI.
Intel needed to enable marketers to find targeted answers
and perform their tasks more efficiently, focusing on the
process for Onboarding Agency Partners as the prioritized
use case.
Intel partnered with Enterprise Knowledge (EK) to validate
Intel’s semantic search priorities, and jumpstart content
preparedness and AI solution development for operational
efficiency.
By the Numbers
What We Did
7
AI Capabilities
Identified
34
Functional &
System
Requirements
3
Functional
Testing
Sessions
126
Document
Chunks
Evaluated
8
Diagnostic
Steps
Average Response Score
(out of 1)
0.83
0.4
After Phase 8 of Testing
Phase 1

WHAT IS A
“GOOD” AI
OUTCOME?

CORRECT
Results are factually accurate and
logically sound. They reflect verified
data and align with the known truth or
accepted standards in the domain. No
misinformation or hallucinations.
COMPLETE
Outputs are comprehensive, addressing
all relevant aspects of the query or task.
CONSISTENT
Results provide stable and repeatable
outputs when given the same inputs in
the same context. Predictability builds
trust and ensures smooth integration
with workflows and systems.
CONTEXTUAL
Outputs are aware of and responsive to
the organization's unique knowledge and
processes. They reflect domain-specific
language, priorities, and current
operational states and landscape.
COMPLIANT
Results align with organizational and
external policies around security,
privacy, access controls, ethical use,
regulatory standards, bias mitigation,
and brand integrity.
The Five C’s to evaluating AI results, reliability, and adoption:
Evaluating AI Readiness

AI For What Purpose?
Intelligent
Retrieval
Search
Summary
Task Support
Intelligent
Checkpoints
Q&A

What is it?
Retrieving the right set of
documents or content units.
Intel Example
How AI Helps
Instead of providing excessive
resources, the AI returns the single,
most relevant document to your role,
region, and business unit.
Intelligent Retrieval
Ask: How do I onboard an agency
partner?
Returns: Link to documents which
describes how to onboard an agency
partner
Intelligent Retrieval: Five C’s

Q&A
What is it?
Asking a question in natural
language and getting an answer
in natural language. Ask: How do I onboard an agency
partner?
Returns: Respond with the speciﬁc chunk
of content that answers the question and
the link to the full page or document.
Intel Example
How AI Helps
The AI pulls the exact passage that
explains how to proceed, rewrites or
highlights it in plain English, and
includes a link to the full document.

Search Summary
partner?
Returns: Complete summary of the
process to onboard an agency partner,
compiled with information across multiple
pages and documents.
How AI Helps
The AI pulls out the steps, stitches
them together into a clear,
end-to-end checklist, and removes
duplicate or conﬂicting guidance.
What is it?
Create a summary of a process
where the process is spread out
across multiple content objects.
Intel Example

Task Support
partner?
Returns: Summary of the process
complete with all of the web forms that
need to be completed in order to execute
the process. Some fields in the web forms
may be pre-filled.
How AI Helps
The AI reads all relevant policies and
procedures, breaks them into clear
steps, and then surfaces the right
web forms at the right time in the
workflow.
What is it?
Not just helping you find the
information, but also helping you
to complete the task.
Intel Example

Intelligent Checkpoints
partner?
Returns: A short “readiness report” for
onboarding the agency partner that: (1) Lists each
required checkpoints; (2) Shows pass / fail for each;
and (3) Includes links back to the speciﬁc systems
or documents needed to close the gaps.
What is it?
Surfacing criteria, ensuring output
meets the criteria, returning output
certifying compliance.
Intel Example
How AI Helps
The AI parses the onboarding
policies and turns them into clear
checkpoints for users.

WHAT MAKES
CONTENT “AI
READY”?

Remove duplicate content through automation.
2. Redundancy
1. Currency
3. Relevancy
5. Structure
6. Semantics
Narrow down outdated content to the
newest / most recent version of
content.
Ensure correctness and reliability of content.
Align content necessary and applicable for day-to-day
operations and speciﬁc use cases through subject
matter experts.
Enrich content with semantics (metadata, taxonomies & business glossary)
Technology
Driven
Human
Driven
AI Ready Content:
New, reliable, enriched with embedded organizational context
Approach
Approach:
Automation-accelerated
clean-up that decreases the
manual burden
Outcomes: Reduction of large
corpus of outdated and
redundant content for AI
consumption
Approach:
Transform unstructured content into
structured content by applying
taxonomy tags and imposing a content
model.
Outcomes: Reliable content that is
validated for relevancy and usability by
experts and is both human-readable
and machine-readable for
incorporation in AI solutions
4. Accuracy
Enables composition and contextualization of content

This line graph
demonstrates the
steady rise in
response quality as
content is
structured and
enriched with
additional
semantic context.
Step 2:
Baseline
Capture
Step 5:
Page
Structure
Awareness
Step 6:
Page
Structure
Optimization
Step 7:
Taxonomy
Enrichment
Step 8:
Graph -
Ontology &
Semantic
Enrichment
Original
answer was
missing roles.
Progression of Response Quality

Compare the battery speciﬁcations between 2024 and 2025 Voltz Sedan safety data sheets
Example 1: Structured Content
TARGET QUESTION
5. Structure Enables composition and contextualization of content

In Practice: Without Structured Content
Real GPT 5
Generation

In Practice: Without Structured Content
RAG-generated
answer is
incomplete,
reﬂecting
limited inter-
and
intra-document
awareness.
Real GPT 5
Generation
“Most relevant”
chunk does not
even contain
the most
relevant
information
being retrieved.

In Practice: With Structured Content
Real GPT 5
Generation

In Practice: With Structured Content
Answers
question across
successive
versions of a
component
from multiple
documents after
identifying the
crucial
difference
Text level
validation
enabling
high-grade
explainability
Real GPT 5
Generation

Example 2: Taxonomy
TARGET QUESTION
6. Semantics Enrich content with semantics (metadata, taxonomies & business glossary)
What does ‘Trip Planner’ refer to?

In Practice: Without Taxonomy
Real GPT 5
Generation

In Practice: Without Taxonomy
Generic
response, lacks
context and a
speciﬁc
relationship to
the ‘Trip
Planner’ term
Real GPT 5
Generation

In Practice: RAG Without Taxonomy
Real GPT 5
Generation

In Practice: RAG Without Taxonomy
Real GPT 5
Generation
Misrepresents Trip Planner as
a current, Pulse X-speciﬁc
“enhanced” feature; invents
extra behaviors (routes, trafﬁc,
auto charging-stop
suggestions).

In Practice: RAG With Taxonomy
Real GPT 5
Generation

In Practice: With Taxonomy
Direct and
relevant
answer
Able to directly
trace the
response and
term back to
the knowledge
base
Real GPT 5
Generation

Example 3: Graph/Ontology
TARGET QUESTION
6. Semantics Enrich content with semantics (metadata, taxonomies & business glossary)
Show me the 2025 Pulse X specs.

In Practice: Without Graph/Ontology
Real GPT 5
Generation

In Practice: Without Graph/Ontology
Due to lack of context, user
has to provide more info
Doesn’t process the year
queried
Generic
response, lacks
speciﬁc
attribute and
relationship
details
Real GPT 5
Generation

In Practice: RAG Without Graph/Ontology
Real GPT 5
Generation

In Practice: RAG Without Graph/Ontology
Real GPT 5
Generation
This answer correctly
preserves some specs but
misstates the drivetrain and
power and drops key
structured ﬁelds like seating
capacity and price (even when
mentioned in source).

In Practice: With Graph/Ontology
Real GPT 5
Generation

In Practice: With Graph/Ontology
Directly
answers query
with a
relevant,
complete
answer
Finds terms
from
knowledge
graph, this
helps provide
context
Real GPT 5
Generation

Transforming Content Operations in the Age of AI

More Related Content

Similar to Transforming Content Operations in the Age of AI

More from Enterprise Knowledge

Recently uploaded

Transforming Content Operations in the Age of AI