Agenda
Text Analytics Forum is a featured event at KMWorld 2026. See the Advance Program PDF.
- Monday Nov 16
- Tuesday Nov 17
- Wednesday Nov 18
- Thursday Nov 19
Optional Pre-Conference Workshops/Conference Day
Optional Conference Day
Sessions
Sessions
9:00 AM
Optional Pre-Conference Workshops/Conference Day
Length: 7 Hour 30 Minutes
Description: Text Analytics Forum is part of KMWorld 2026 featuring five co-located events: KMWorld 2026, Taxonomy Boot Camp, Enterprise Search & Discovery, Text Analytics Forum, and Enterprise AI World. Upgrade to a Platinum Pass for your choice of two preconference workshops or access to Taxonomy Boot Camp on Monday. Workshops are also separately priced.
5:00 PM
Optional Pre-Conference Workshops/Conference Day
Length: 1 Hour 30 Minutes
Description: Join us for the opening of the Enterprise Solutions Showcase to explore the marketplace and connect with the community. Discover cutting-edge products and services from the industry’s top companies while enjoying live music, a complimentary beer and wine bar, and a selection of light hors d'oeuvres. Whether you're looking to scope out new tech or reconnect with industry peers, this vibrant reception is the perfect place to build your network.
7:00 PM
Optional Pre-Conference Workshops/Conference Day
Length: 2 Hours
Description: Join us for a night of friendly competition! Whether you're a seasoned strategist or just looking to unwind, we'll have a variety of games ready to play. Come grab a seat, meet fellow attendees, and let the games begin.
8:30 AM
Optional Conference Day
Length: 8 Hour 30 Minutes
Description: Text Analytics Forum is part of KMWorld 2026. Upgrade to a Platinum or Gold Pass for extended access to KMWorld 2026, Enterprise Search & Discovery, Taxonomy Boot Camp, and Enterprise AI World, a series of co-located events happening alongside Text Analytics Forum 2026. See the registration page for details.
5:00 PM
Optional Conference Day
Length: 1 Hour
Description: Wind down after a full day of stimulating sessions with a casual happy hour right on the showcase floor. Grab a drink, visit the booths, and dive deeper into conversations with fellow attendees, speakers, and sponsors. It’s the perfect, laid-back setting to talk shop, swap ideas, and solidify your new connections.
8:00 AM
8:30 AM
Keynotes
Length: 1 Hour
Speaker(s):
Kim Glover, Director, TechnipFMC
Julie Mohr, Principal Analyst, Forrester
Maggie Starkey, Knowledge Management Specialist, Kraton Corporation
Elizabeth Turner, Founder & Principal, Turner Adaptive Strategies
Kim Glover, Director, TechnipFMC
Julie Mohr, Principal Analyst, Forrester
Maggie Starkey, Knowledge Management Specialist, Kraton Corporation
Elizabeth Turner, Founder & Principal, Turner Adaptive Strategies
Description: In today’s environment of constant disruption, change is no longer a one-time event—it is continuous, complex, and deeply human. Our panel discusses what it takes to lead transformation while conditions are still evolving and draws on real-world experience across industries and organizational contexts. The conversation looks at how to support people through uncertainty, equip leaders and employees to communicate with clarity and confidence, and design change approaches that teams can realistically absorb. Panelists focus on the role of KM as a practical enabler of change—helping preserve continuity, reduce risk, and maintain momentum during periods of transition. They offer candid insights, practical lessons learned, and tips on dealing with change in a world that refuses to stand still. Get lots of ideas on how to navigate continuous change by connecting people, knowledge, and action.
9:30 AM
Keynotes
Length: 15 Minutes
Speaker(s):
Philip Brittan, CEO, Bloomfire
Philip Brittan, CEO, Bloomfire
Description: Knowledge repositories were built to store information, but storing information and activating organizational intelligence in an AI-first world are fundamentally different problems. Brittan makes the case that the next evolution in KM is context: building a persistent, measurable understanding of what an organization knows, who knows it, and whether that knowledge is healthy enough to be trusted. Drawing on Bloomfire’s innovative work building an enterprise intelligence engine, he discusses the power of conducting knowledge health assessments with enterprise customers; explores what it means to move from passive content management and RAG-based search to an active, context-rich intelligence infrastructure; and shares real-world examples of organizations building intelligence infrastructures. Get tips on creating healthier knowledgebases, reductions in compute power, and overall better AI outcomes.
9:45 AM
Keynotes
Length: 15 Minutes
Speaker(s):
Rick Nucci, Co-Founder and CEO, Guru
Rick Nucci, Co-Founder and CEO, Guru
Description: Every company wants AI. Almost none have solved the thing that makes AI trustworthy: the knowledge underneath it. Research shows most organizations can manually verify only 8%–12% of their knowledge, and AI scales the errors in everything else at machine speed. This is the knowledge accuracy gap, and it is the real reason AI programs stall. Nucci discusses how KM is not a support function but rather the infrastructure that decides whether AI works at scale or quietly fails. He looks at why accuracy, not model choice, is the true bottleneck; what separates the companies pulling ahead; and why knowledge managers are about to become the most important people in their organizations. If you own knowledge at your company, this is the year that work stops being invisible.
10:00 AM
10:45 AM
Sessions
Length: 45 Minutes
Speaker(s):
Tom Reamy, CEO, KAPS Group
Tom Reamy, CEO, KAPS Group
Description: What are the current and future trends for the field of text analytics? Join program chair, Tom Reamy, for an overview of the conference themes and highlights and a look at what is driving the field forward. The theme this year is Text Analytics (TA) and the Integrated Enterprise.
Unstructured text is everywhere, and, despite multiple and varied attempts to get real value from all that text, it continues to be a problem child. Any solution requires an integrated approach to utilize all the varied types and uses of unstructured text. Text analytics (TA) is a critical component for all efforts to utilize unstructured text in the enterprise and is the key to a successful integration of this essential enterprise resource. This year’s conference explores how TA can enhance and be enhanced by four other key players: knowledge organization (taxonomy and knowledge graphs), search and search-based applications, knowledge management (KM) and learning, and, of course, GenAI and machine learning.
11:45 AM
Sessions
Length: 45 Minutes
Speaker(s):
Christopher Davidson, Founder, DeerTrail Innovation Group
Doug Beeson, Ontologist, Semantic Arts
Christopher Davidson, Founder, DeerTrail Innovation Group
Doug Beeson, Ontologist, Semantic Arts
Title: Building Scientific Latent Knowledge Graphs
Time: 11:45 AM - 12:30 PM
Description: Similarity-based knowledge graphs have a precision problem. A cosine distance edge between two entities tells you they appear near each other in the text. It doesn't tell you the direction, the mechanism, or the evidence. For high-stakes domains—drug discovery, regulatory review, competitive intelligence—that gap has real consequences. A graph that can't distinguish whether a compound causes an adverse event from one that treats it isn't a knowledge graph. It's an indexed co-occurrence matrix. This session introduces Epistract, an open source agentic architecture for biomedical knowledge graph construction that replaces similarity-based assembly with comprehension-based extraction. Instead of embedding chunks and measuring distances, parallel LLM agents read entire documents with domain understanding and produce structured JSON conforming to a biomedical schema: 17 entity types, 30 typed directional relation types, grounded in 40-plus established ontologies. Every extracted relationship carries the source passage, a confidence score, and ontology grounding. Molecular identifiers are validated deterministically—not approximated. The system was evaluated across six drug discovery domains with zero retraining: KRAS G12C inhibitor landscape, immuno-oncology combinations, Alzheimer's target validation, rare disease therapeutics, cardiovascular inflammation, and GLP-1 competitive intelligence across PubMed, Google Scholar, and patent corpora. The session presents the architecture, the evaluation methodology, and the broader design principle: For knowledge-intensive domains, text analytics needs to move from proximity to comprehension. It is based on the paper “Beyond RAG: Domain-Specific Agentic Architecture for Biomedical Knowledge Graph Construction” (bioRxiv, DOI pending).
Title: Graph Patterns for Unstructured Data Extraction
Time: 11:45 AM - 12:30 PM
Description: This talk introduces RDF graph patterns, a new and effective way to extract knowledge graph data from unstructured text. Graph patterns grew from RDF, OWL, and SPARQL, languages designed for the representation of data as facts and knowledge. A pattern stands in for the CONSTRUCT portion of a SPARQL query. The WHERE portion is handled by the LLM. It decides how each pattern triple gets bound, relying on a context that is typically already loaded with the ontology and taxonomy as RDF triples. Semantic Arts has created a new vocabulary for these patterns.
12:30 PM
Attendee Lunch in the Enterprise Solutions Showcase
Length: 1 Hour
1:30 PM
Sessions
Length: 45 Minutes
Speaker(s):
Jans Aasman, CEO, Franz Inc
Jans Aasman, CEO, Franz Inc
Description: For the last 2 years, many organizations have acted as if LLMs alone could solve text analytics. They cannot. LLMs are impressive at reading, summarizing, and extracting from documents. But enterprise text intelligence requires more than plausible answers. It requires entities that resolve correctly, relationships that can be queried, claims that can be traced back to source documents, contradictions that can be detected, and analytics that work across thousands or millions of records. That is where knowledge graphs stop being optional. GraphRAG was the first wave of rediscovery: People realized that LLMs need better grounding. But GraphRAG is only an intermediate step. The real architecture combines LLMs, ontologies, vocabularies, knowledge graphs, vector search, rules, and reasoners into a neuro-symbolic stack. This makes text not merely searchable, but computable, explainable, and auditable. This session shows how this works in practice across several real-world domains: invoice and contract compliance, financial anomaly detection, clinical trial matching, patient-event knowledge graphs, and an FAA maintenance knowledge graph that harmonizes aircraft maintenance records. These examples demonstrate how ontology-driven extraction, graph-based normalization, aggregate analytics, and symbolic reasoning can dramatically increase the intelligence extracted from messy enterprise text. The core argument is simple: If your text analytics system cannot build and reason over a knowledge graph, it is not really doing text intelligence. It is doing advanced document guessing.
2:30 PM
Sessions
Length: 45 Minutes
Speaker(s):
Alice Chung, Field Medical Excellence and Enablement Lead, Genentech
Tom Sabo, Advisory Solutions Architect, SAS
Alice Chung, Field Medical Excellence and Enablement Lead, Genentech
Tom Sabo, Advisory Solutions Architect, SAS
Title: Agentic Knowledge Ecosystems: Modeling the Convergence of Hybrid Search, Active Metadata, & Self-Improving AI Pipelines
Time: 2:30 PM - 3:15 PM
Description: The dream of the integrated enterprise has moved past standard chat search engines into autonomous systems that look up information, execute actions, and evaluate their own logic to continuously improve. However, corporate implementations at times struggle with inaccurate responses, information security gaps, and fragile prompt rules that render generative applications unreliable for critical workflows. This presentation shifts away from low-level coding setups to focus on the essential conceptual modeling blocks needed to build a robust, self-correcting corporate knowledge platform. We look closely at how text moves through an enterprise workflow to outline the trade-offs between baseline vector search, high-precision hybrid search, and hierarchical graphs. Attendees explore how basic text matching falls short on exact business terms, why combining keyword matching with meaning maps protects technical accuracy, and how relational graph structures keep multistep reasoning anchored. Moving along the workflow, the session demonstrates how to design automated workflows where smart agents use active metadata tags to secure data, eliminate entitlement drift at query time, and pull live database insights via Table-Augmented Generation (TAG). Finally, we explore the conceptual modeling of self-improving data loops, where an AI system analyzes its own mistakes using automated feedback to update its own core prompts and tools. Attendees leave with a practical conceptual blueprint to scale their data engines from basic answer engines into secure, self-evolving organizational assets.
Title: Text: The Other Synthetic Data
Time: 2:30 PM - 3:15 PM
Description: Synthetic data isn’t just rows and columns. Organizations increasingly need safe, realistic text to accelerate decision making while protecting privacy. This session explores how to generate high-utility synthetic text for sensitive sources like customer interactions, medical notes, and police reports. Using text analytics and GenAI, Sabo demonstrates an iterative, cross-domain framework that blends new and historical snippets to produce governed synthetic narratives. See how to develop an AI Agent to construct targeted LLM prompts and how these capabilities help you unlock insights, improve model performance, and meet strict privacy and compliance requirements.
3:15 PM
Coffee & Networking Break Coffee & Networking Break in the Enterprise Solutions Showcase
Length: 45 Minutes
4:00 PM
Sessions
Length: 1 Hour
Speaker(s):
Joseph Hilger, COO, Enterprise Knowledge
Rachel Carrier, Senior Technical Analyst, Enterprise Knowledge
Sara Nash, Practice Lead, Semantic Engineering and AI, Enterprise Knowledge
Ben Cross, Technical Consultant, Enterprise Knowledge
Joseph Hilger, COO, Enterprise Knowledge
Rachel Carrier, Senior Technical Analyst, Enterprise Knowledge
Sara Nash, Practice Lead, Semantic Engineering and AI, Enterprise Knowledge
Ben Cross, Technical Consultant, Enterprise Knowledge
Title: Managing Risk & Financial Services With AI & Semantics: A Practical Case Study
Time: 4:00 PM - 5:00 PM
Description: Financial services organizations are under increasing pressure to improve how critical knowledge is governed, discovered, and operationalized across complex and rapidly evolving business environments. Yet many continue to face challenges driven by fragmented content ecosystems, large unwieldy datasets, inconsistent metadata, and search experiences that lack contextual relevance. This session explores how semantic technologies, GenAI, linked data, and knowledge graph capabilities can be combined to create a scalable, data-centric foundation for discovering contractual risks across large legal datasets. Drawing from real-world implementation experience within financial services, our speakers discuss how ontology development and semantic-enabled AI pipelines can condense large volumes of free-text risk descriptions into standardized risk concepts, reducing duplication and inconsistency while improving the ability to analyze, govern, and operationalize risk information across the enterprise. The session walks through key phases of implementation, from strategy and architecture design to implementation, governance, and enterprise scale, while highlighting lessons learned and practical considerations for operationalizing semantic capabilities within highly regulated organizations.
Title: Connecting Structured & Unstructured Data Through Semantic Layers to Power AI Platforms
Time: 4:00 PM - 5:00 PM
Description: Financial firms rely on both structured data (e.g., deal records, portfolio data, CRM systems) and unstructured content (e.g., diligence reports, investment memos, emails) to make critical investment decisions. However, the context needed to interpret structured data often resides in unstructured content, making the connection between the two a major barrier to effective decision making. This challenge is compounded by organizational silos, where data warehousing and content management teams operate separately. Without a unifying layer, insights remain fragmented, slowing analysis and limiting the reuse of institutional knowledge. Semantic layers address this by linking structured and unstructured data through shared, business-oriented entities such as companies, sectors, transactions, and experts. By enabling unstructured data to enrich structured data with context—and structured data to organize and scale unstructured analysis—firms can leverage both in concert. This approach not only surfaces context-rich insights and accelerates due diligence, it also aligns teams around a common language and data model. In this session, EK’s Nash and Cross share how two leading sovereign wealth funds partnered with EK to design and scale a semantic-enabled platform. Attendees learn how to design scalable semantic architectures that connect structured and unstructured data and how to align teams, governance, and workflows to support long-term adoption and enterprise-wide impact.
5:30 PM
Sessions
Length: 1 Hour
Description: Connect with peers over drinks and great conversation on KM, AI, search, data analytics, and taxonomies. Exchange insights and challenges in a relaxed atmosphere at the hotel bar. (This is a cash bar event.)
8:00 AM
8:30 AM
Keynotes
Length: 45 Minutes
Speaker(s):
Rob Stein, Keynote Speaker, Consultant, Executive Coach, Author, Author, Impossible to Fail: The Step-by-Step Formula to Guarantee Your Success in Anything
Rob Stein, Keynote Speaker, Consultant, Executive Coach, Author, Author, Impossible to Fail: The Step-by-Step Formula to Guarantee Your Success in Anything
Description: Rob Stein believes success becomes far more attainable when ambition is paired with the right blueprint, decisive action, relentless consistency, and time. A national speaker, performance coach, entrepreneur, and author, Stein has built and scaled multiple businesses, created the Impossible to Fail framework, and coached thousands of entrepreneurs and high performers on how to move from overthinking to focused execution. In this session, Stein shares a clear, repeatable approach for turning big goals into sustainable progress. Attendees learn how to overcome fear, simplify decision making, build traction, and move forward with clarity and confidence. Drawing from the principles in Impossible to Fail, Stein offers practical strategies for leaders and teams who need less noise, more focus, and a reliable way to take action without hype, fluff, or guesswork.
9:15 AM
Keynotes
Length: 15 Minutes
Speaker(s):
Fahreen Kurji, Chief Customer Intelligence Officer, Behavox
Fahreen Kurji, Chief Customer Intelligence Officer, Behavox
Description: As regulators raise the bar on explainability and accountability, compliance teams are under pressure to move beyond simply generating alerts toward building systems that can withstand scrutiny. Our popular speaker discusses what "defensibility" means in practice for AI-driven surveillance; the shift from detection to demonstrable, auditable control; and lessons learned from deploying AI across financial crime compliance programs. Using real-world examples, she provides insights, ideas, and tips for the successful enterprise.
9:30 AM
Keynotes
Length: 15 Minutes
Speaker(s):
Stephane Donze, Founder & CEO, AODocs
Stephane Donze, Founder & CEO, AODocs
Description: Autonomous AI agents change the way we access information, interact with enterprise data, and decide between building or buying our applications. These agents also come with many risks and unknowns, especially when applied to business-critical information. Donzé discusses how to ensure conversational AI agents provide reliable, hallucination-free answers; how to avoid chaos when your business users can vibe-code their own apps in a few hours; and what kind of enterprise foundation you need to fully benefit from AI acceleration while controlling the risks. He illustrates with real-world examples and provides lots of insights and tips.
9:45 AM
Keynotes
Length: 15 Minutes
Speaker(s):
Jan Stihec, Director, Data & GenAI, Shelf
Jan Stihec, Director, Data & GenAI, Shelf
Description: As organizations race to adopt AI and agentic AI, much of the attention is focused on models, workflows, and interfaces. But the long-term success of AI will depend on something more fundamental: the structure, interpretability, quality, and accessibility of the organization’s knowledge and data. AI systems cannot reason effectively, act reliably, or earn trust if they cannot understand the information they depend on. Enterprise knowledge is often fragmented across documents, systems, teams, taxonomies, and processes. For AI agents to deliver value beyond narrow use cases, they must be able to navigate this complexity, interpret context correctly, retrieve the right information, and apply it in ways that are accurate, explainable, and aligned with business needs. Stihec discusses why knowledge and data must become the foundation of a scalable AI strategy; looks at the gap many organizations face when they invest heavily in AI models and workflows while underinvesting in the knowledge layer that makes those systems useful; and provides a knowledge-first approach that can improve trust, reduce hallucinations, expand AI use cases, and create the conditions for AI and agents to scale across the enterprise. Illustrating with real-world examples, he shares insights and ideas for the successful enterprise.
10:00 AM
Coffee & Networking Break in the Enterprise Solutions Showcase
Length: 15 Minutes
10:15 AM
Sessions
Length: 45 Minutes
Speaker(s):
Duane Degler, Principal, Design for Context
Duane Degler, Principal, Design for Context
Description: Text analysis, content structure analysis and GenAI summarization have come a long way. They can be very effective in search applications for long-form documents and even scanned archival materials. One challenge users face is looking across very large collections of documents, where the goal is not finding one document, but seeing the patterns of their interest across a large number of documents and images. This is a common task for research, and a starting point for new content authoring. As a case study, archival collections hold the stories of people, objects, products, events, and histories. These are contexts, across time in many cases, with long histories and hidden insights. Users who want to create new stories or understand complex events don’t want the needle in the haystack, they want to find enough needles and threads to weave new, relevant stories. This talk describes work done on pattern-finding, density-mapping, and creating new exploratory interface concepts that help users look across very large content collections to see how their subjects of particular interest appear across the landscape of the source material. This offers wayfinding that fuels creative research and discoveries. Techniques in text and entity analysis help provide a signal strength for exploring across source materials in ways that can help users utilizing those signals as starting points for synthesizing new content, honing interpretations, and surfacing insights from the historic story landscape.
11:15 AM
Sessions
Length: 45 Minutes
Speaker(s):
Lauren Clark Hill, Expert Solutions Engineer, Squirro
Urmi Majumder, Principal Consultant, Enterprise Knowledge
Kyle Garcia, Technical Consultant, Enterprise Knowledge
Lauren Clark Hill, Expert Solutions Engineer, Squirro
Urmi Majumder, Principal Consultant, Enterprise Knowledge
Kyle Garcia, Technical Consultant, Enterprise Knowledge
Title: Governing Known & Novel Entities to Reduce Hallucinations in GenAI
Time: 11:15 AM - 12:00 PM
Description: As organizations adopt GraphRAG and LLM-based applications to make sense of unstructured content, many are discovering a familiar problem in a new form: Meaning breaks down when governance of terminology, structure, and relationships is inconsistent. What is often labeled as a GenAI hallucination is, in practice, frequently a downstream symptom of weak or uneven control over how enterprise concepts are defined and applied. This session approaches the problem from a taxonomy-first perspective. Rather than focusing on text analytics as a standalone discipline, it examines how taxonomy decisions directly shape the quality of entity extraction, retrieval, and, ultimately, GenAI outputs. A key emerging challenge is the way LLMs extract and mix two types of entities from unstructured text: known entities, which are aligned to governed taxonomies and knowledge graphs, and novel entities, which are inferred dynamically from context. Without clear governance boundaries between the two, GraphRAG systems can unintentionally combine stable enterprise concepts with transient or ambiguous ones—producing answers that are fluent and well-supported, but semantically incorrect. This session focuses on practical implications for taxonomy practitioners that include how taxonomy structure influences entity extraction quality in LLM-driven systems, where synonym control and concept definitions directly impact hallucination risk, why “novel entity discovery” can quietly undermine governed vocabularies, and how GraphRAG retrieval amplifies small taxonomy inconsistencies into visible AI errors. Rather than treating text analytics as a separate technical domain, the discussion reframes it as an extension of taxonomy governance into unstructured environments. The goal is to show how taxonomy professionals can extend their existing practices to better govern LLM-driven entity extraction and reduce hallucination risk in GenAI systems. Attendees leave with a clearer understanding of how traditional taxonomy work now directly shapes the reliability of modern AI outputs and where their governance role becomes essential in preventing semantic drift at scale.
Title: How to Augment Taxonomy Change Management With Statistics
Time: 11:15 AM - 12:00 PM
Description: How do organizations effectively determine when it’s time for a taxonomy update? How much content change is too much? With ever-changing corpuses of content, it can be difficult to determine if a taxonomy is still relevant and applicable, especially in auto-tagging applications. As such, taxonomists, information architects, and semantic engineers alike benefit greatly from referenceable datapoints that they can use to decide on the best course of action when considering taxonomy change against a content corpus. While there are multiple statistical approaches to consider with varying degrees of complexity, it can be difficult to reconcile them and select the correct methods for given use cases. Together, EK’s Majumder and Garcia share various automated methods to measuring taxonomy applicability to a content corpus, statistical measures to look for when evaluating auto-tagged content, and the best ways to measure the semantic drift of a content corpus as assets are created, modified, and sunsetted. By measuring semantic drift and evaluating taxonomy fit through automated, data-driven methods, organizations can maintain confidence in their auto-tagging solution without the manual effort this process typically demands. This expertise is grounded in real-world case studies in which EK successfully evaluated multiple taxonomies across numerous organizations for their relevance to content for the purposes of auto-tagging using multiple discrete methods. Attendees of all technical experience levels are welcome to attend.
12:00 PM
Sessions
Length: 1 Hour
Description: KMWorld magazine is proud to sponsor the 2026 KMWorld awards, KM Promise & KM Reality, which are designed to celebrate the success stories of knowledge management. The awards will be presented along with Step Two’s Digital Awards, where you get a sneak peek behind the firewall of these organizations.
1:00 PM
Sessions
Length: 45 Minutes
Title: Finding the Common Denominator Between Your Structured & Unstructured Data
Time: 1:00 PM - 1:45 PM
Description: The schemas of most enterprise applications are far too complex and far too idiosyncratic to be of much use in harmonizing unstructured data. But merely using AI to tag documents doesn’t do much to promote integration of structured and unstructured data. It turns out there is a way to unify the two realms. It involves starting with a semantic core and mapping both the structured and the unstructured to this same core. This talk describes how this process works and review some case studies.
Title: Fixing Enterprise & Agentic AI’s Weak Link
Time: 1:00 PM - 1:45 PM
Description: GenAI and agentic AI are dependent on the quality of their training sets, and when it comes to enterprise data, the situation is grim. As we’ve known for years, enterprise content is filled with badly written and multiple copies and worse, near copies. Trying to manually turn messy enterprise content into a useful training set is a daunting and expensive task filled with individual variation, bias, and neglect. While RAG and GraphRAG are powerful technologies, RAG is dependent on enterprise search which has an abysmal track record. The answer is to utilize text analytics capabilities of auto-categorization, data and fact extraction, and similarity matching to clean up all that content. This talk covers the entire process, from the development of sophisticated auto-categorization rules that incorporate document structure and data extraction rules. These rules are then applied to any and all enterprise content to build high-quality training sets. It concludes with how to utilize those training sets in enterprise GenAI and agentic AI.
2:00 PM
Sessions
Length: 45 Minutes
Speaker(s):
Matt Van Vleet, Managing Director, LiminalArc
Kelsey Young, Principal Technical Advisor, SAS
Matt Van Vleet, Managing Director, LiminalArc
Kelsey Young, Principal Technical Advisor, SAS
Title: Keeping AI Honest: A/B Testing for Agent Reliability & Maturity
Time: 2:00 PM - 2:45 PM
Description: In ecommerce, change relies heavily on experimentation. For most companies, it’s standard operating procedure to test, measure, and A/B test things like user experiences, pricing strategies, and recommendation engines. But in the world of agentic AI, that discipline often gets ignored. Teams deploy agents, tweak the prompts, upgrade the model, and hope for the best. But agents deployed isn’t the goal. The goal is to deploy agents that you actually trust in production. The nuance is subtle, but critical. The reality is that systems built on AI aren’t static; they evolve. Model updates, prompt changes, data drift, and shifting usage patterns make production-ready agents hard to identify without a systematic way to measure their behavior. And teams are left struggling to tell whether their agents are improving or slowly degrading across time. This talk explores a different approach in which agents are treated like products, and experimentation, refinement, and optimization are the norm rather than the exception. By applying the principles of A/B testing, organizations can compare agent behaviors, validate improvements, and detect regressions before they impact their customers. This gives teams the ability to base decisions on concrete evidence rather than leaving them to debate a prompt's efficacy based on intuition.
Title: Text Analytics in Orbit
Time: 2:00 PM - 2:45 PM
Description: Government agencies must innovate with limited budgets and avoid duplicative research while staying ahead of rapidly evolving technologies. Kelsey’s session explores how agencies can use text analytics and GenAI to better understand and optimize technology investments. It also covers how an AI-powered agent analyzes public research and contracting data, enabling interactive exploration, trend analysis, and spending insight across time. Learn how this approach of trusted and traditional text analytics with GenAI supports more informed investment decisions, reduces redundancy, and enables cost-effective, mission-ready innovation for space based national security systems.
3:00 PM
Sessions
Length: 45 Minutes
Speaker(s):
Nathan Cummins, Product Director, AI/ML Services, Press Ganey/Forsta
Shari Maginnis, Distinguished Data Scientist, Press Ganey/Forsta
Nathan Cummins, Product Director, AI/ML Services, Press Ganey/Forsta
Shari Maginnis, Distinguished Data Scientist, Press Ganey/Forsta
Description: GenAI has accelerated exploratory text analysis, but many organizations still struggle to make results consistent, explainable, and aligned with enterprise knowledge structures. For teams grounded in taxonomy, metadata, and search, the challenge is not access to models, but control.
This session presents a real-world case study of building a production-grade text analytics system (Narrative HX) that moves beyond ad hoc prompting toward repeatable, scalable outcomes. Co-presented by a product leader and an applied AI engineer, it focuses on how to systematically incorporate context (domain knowledge, hierarchical taxonomies, and structured metadata) into LLM-driven workflows. Rather than treating LLMs as standalone classifiers, our speakers show how to combine probabilistic outputs with deterministic structures to improve consistency, traceability, and alignment with existing knowledge models. Key topics include using taxonomies and metadata as control layers for classification and labeling; designing few-shot and structured prompts to reduce variance in outputs; evaluating and tuning for consistency, not just accuracy; trade-offs between model flexibility and system-level constraints; and evolving hierarchical models alongside changing business needs. This session emphasizes implementation patterns, failure modes, and design trade-offs, not tools or hype. Attendees leave with practical approaches for integrating GenAI into existing knowledge management systems while maintaining rigor, repeatability, and trust.
4:00 PM
Closing Keynote
Length: 1 Hour
Speaker(s):
Kim Glover, Director, TechnipFMC
Ross Smith, WW Support Leader, AI First, Microsoft
Sandra Montanino, Founder & Principal, Navig8 PD
Brian Pichman, Director, Strategic Innovation, Evolve Project
Beth Rudden, CEO and Chairwoman, Bast.ai
Kim Glover, Director, TechnipFMC
Ross Smith, WW Support Leader, AI First, Microsoft
Sandra Montanino, Founder & Principal, Navig8 PD
Brian Pichman, Director, Strategic Innovation, Evolve Project
Beth Rudden, CEO and Chairwoman, Bast.ai
Description: Where is KM going with all the AI developments for the enterprise? How are our organizations responding to the social structures and changes in our world? How are they innovating and exceeding customer expectations? Hear highlights from the conference speakers and get inspiration and ideas from our practitioners and futurists to be ready for KM in 2027.
