Skip to article
Skip to main content

Hands-on demos of Squirro Neo, GraphRAG and our agents. No sign-up, no sales call. Try Them Now

News

Squirro at KMWorld 2026: GraphRAG, Taxonomy and AI Agents in Production

Kmworld Entrance - Jwg Mariott Wshington

Key Takeaways | TL;DR

  • Squirro exhibits at Booth #301 at KMWorld 2026, November 16 to 19 at the JW Marriott in Washington, DC.
  • CEO Dave Clarke opens the Enterprise Search & Discovery keynote block on Tuesday, November 17 at 9:30am with "The Amnesia Problem: Why Enterprise KM Forgets Everything It Learns".
  • Expert Solutions Engineer Lauren Clark Hill runs two sessions on GraphRAG failure and on governing entities to reduce hallucinations.
  • At the booth you can run live demos of GraphRAG, taxonomy management and production-ready AI agents, or try them now in the Demo Center with no sign-up.
  • Not registered yet? Book a meeting with us and we can help with a discount on your pass.

KMWorld 2026: GraphRAG, Taxonomy and AI Agents in Production

A field engineer asks the knowledge base for the right pressure setting. The answer comes back fast, sounds confident and applies to a different machine configuration. Nobody notices until the service visit goes wrong.

That is the problem most of KMWorld 2026 will be talking about. The model is no longer the bottleneck. The knowledge layer underneath it is: how well it is structured, governed and grounded in what the organization actually knows.

Squirro is a Diamond sponsor of KMWorld 2026 in Washington, DC. Come to Booth #301 to see how the enterprise AI context platform finds the right knowledge, structures it with taxonomies and knowledge graphs, and runs agents that act on it, in production, with every answer traceable.

KMWorld 2026 at a glance

 

Dates

Monday, November 16 to Thursday, November 19, 2026

Venue

JW Marriott Washington DC, 1331 Pennsylvania Avenue NW

Squirro booth

#301, Showcase floor

Squirro sessions

CEO keynote (Tue 9:30am) and two Taxonomy and Text Analytics sessions

Co-located events

KMWorld, Taxonomy Boot Camp, Enterprise Search & Discovery, Text Analytics Forum, Enterprise AI World

Book a meeting at KMworld

The knowledge graph that GraphRAG draws on is one part of a semantic layer: an organization’s shared model of its business, covering what its terms mean, how they relate, and which rules govern them. GraphRAG is how retrieval puts that model to work. For the wider picture, see What Is a Semantic Layer, and When Does GenAI Need One?

Which Sessions Is Squirro Presenting at KMWorld 2026?

Squirro speaks on Tuesday and Wednesday, across Enterprise Search & Discovery, Taxonomy Boot Camp and Text Analytics Forum. The three sessions make the same argument from three angles: AI is only as good as the knowledge structure it works on.

Session

Speaker

When

Track

The Amnesia Problem: Why Enterprise KM Forgets Everything It Learns

Dave Clarke, CEO

Tue, Nov 17, 9:30 to 9:45am

Enterprise Search & Discovery, opening keynote

Knowledge Heist! An Interactive Investigation of Failure in GraphRAG Systems

Lauren Clark Hill, Expert Solutions Engineer

Tue, Nov 17, 2:45 to 3:30pm

Taxonomy Boot Camp, Track 1: Semantics in the Age of AI

Governing Known & Novel Entities to Reduce Hallucinations in GenAI

Lauren Clark Hill, Expert Solutions Engineer

Thu, Nov 18, 11:15am to 12:00pm

Text Analytics Forum: AI, Text Analytics, & Taxonomies

The Amnesia Problem: Why Enterprise KM Forgets Everything It Learns

Most enterprise AI is stateless. Each interaction begins from scratch, expertise evaporates between sessions, and the 80% of institutional knowledge that lives as tacit judgment in expert intuition, cultural norms, and unwritten heuristics—remains entirely invisible to the system. Dave's take: the leaders of the next decade won't have the best model; they will be those who build the best memory.

Knowledge Heist! An Interactive Investigation of Failure in GraphRAG Systems

This is an investigation, not a lecture. Attendees reconstruct how a GraphRAG system produced a confident but wrong answer, and trace it back through taxonomy design, tagging and graph relationship decisions. Bring your questions to Booth #301 afterwards: the governance layer that prevents this failure is what we demo.

Governing Known & Novel Entities to Reduce Hallucinations in GenAI

What gets labeled a GenAI hallucination is often a symptom of inconsistent taxonomy governance. Lauren Clark Hill walks through the difference between known entities, aligned to a governed taxonomy, and novel entities inferred on the fly during LLM extraction, and shows how to govern both.

What Will You See at Squirro Booth #301?

 

Every keynote this week will explain why ontology matters. At Booth #301 we show it working. Our team runs live demos on Squirro Neo, the platform that brings enterprise AI search, the semantic layer and AI agents together:

  • GraphRAG, with and without the graph. The same pricing question asked twice: once with plain retrieval, once grounded in a knowledge graph. You see where the first answer goes wrong and why the second one holds.
  • Squirro Graphite. Taxonomy, ontology and knowledge graph management, the layer that tells AI which concept, product or configuration a question refers to.
  • Squirro Classifier. Automated classification of content against your own taxonomy, with humans reviewing where it matters.
  • AI agents in production. Agents from the Squirro Agent Catalog that act on governed knowledge, not just answer questions: the Squirro Instaquote Agent turns a spare-part email into an SAP quotation with a human approval step, the R&D Knowledge Discovery and Deduplication Agent checks a planned experiment against past ones, and the RFP and Proposal Response Agent drafts cited answers from winning proposals.
  • Knowledge management, manufacturing and finance scenarios on Squirro Neo, built around the questions KM teams get every day.

Can't wait until November 17? Most of our booth demos are live in the Squirro Demo Center: 13 interactive demos on sample data, no sign-up and no sales call required. Try one before the show, then come to the booth with the question it did not answer.

How Did a Global Advanced Manufacturer Reach 94% Accuracy?

 

One of the case studies we bring to Booth #301 starts with a familiar KM problem. A global advanced manufacturing company had decades of technical knowledge in service documentation and service history. Its knowledge platform was one of the most used tools in the business, with 16,000 users. But it could not reliably tell which of many product configurations a field engineer's question referred to, and a setting that is right for one configuration is wrong for another.

With other systems, the team struggled to get accuracy above 70%. With Squirro, on the same content, the customer measured 94% accuracy on technical troubleshooting queries in its latest round of testing. The remaining 6% of answers were rated neutral.

The gain did not come from a better model. It came from tuning five elements together: orchestration, taxonomy and ontology, guardrails, governance and retrieval architecture. Squirro Graphite encodes which components are shared across configurations and which document governs each setup. The engineer on site stays the decision maker, and the system flags its own uncertainty. The customer's internal AI team, the group most skeptical of LLMs, approved the system after evaluation. In the words of its Managing Director, he did not know this level of accuracy was achievable.

Read the advanced manufacturing case study

The same pattern runs in production elsewhere. Bühler, the Swiss manufacturing giant, runs Squirro agents in production, starting with quoting. ZwickRoell, the materials testing specialist, uses Squirro to make 160 years of institutional knowledge searchable.

Book a Meeting With Squirro at KMWorld 2026

The showcase floor gets busy between sessions. Book a slot with our team in advance and we will run the demo on your use case: one repetitive but complex process where accuracy matters, and what it would take to get it into production. Thirty minutes at the booth is enough to see whether your process is the right starting point.

Not registered for KMWorld 2026 yet? Let us know when you book and we can help you get a discount on your pass.

Book a meeting at KMWorld 2026

Frequently Asked Questions

 

When and where is KMWorld 2026? KMWorld 2026 runs from November 16 to 19, 2026 at the JW Marriott Washington DC, 1331 Pennsylvania Avenue NW. It brings together five co-located events: KMWorld, Taxonomy Boot Camp, Enterprise Search & Discovery, Text Analytics Forum and Enterprise AI World.

Where is Squirro's booth at KMWorld 2026? Squirro is at Booth #301 on the Showcase floor. The showcase opens with a reception on Monday, November 16 from 5:00 to 6:30pm, then runs Tuesday from 10:00am to 6:00pm and Wednesday from 10:00am to 4:00pm.

Which sessions is Squirro presenting? CEO Dave Clarke gives "The Amnesia Problem: Why Enterprise KM Forgets Everything It Learns" on Tuesday at 9:30am in the Enterprise Search & Discovery keynote block. Lauren Clark Hill presents "Knowledge Heist!" on Tuesday at 2:45pm at Taxonomy Boot Camp and "Governing Known & Novel Entities to Reduce Hallucinations in GenAI" on Wednesday at 11:15am at Text Analytics Forum.

What is GraphRAG and why does it matter for knowledge management? GraphRAG grounds generative AI answers in a knowledge graph, not only in text chunks retrieved by similarity. The graph tells the system how concepts, products and documents relate, so it can pick the right answer when several look alike. For KM teams, it turns existing taxonomies and ontologies into the accuracy layer for AI.

Can I try Squirro's demos before KMWorld? Yes. The Squirro Demo Center hosts 13 interactive agent demos on sample data, with no sign-up required. The same demos run at Booth #301.

Can I get a discount on a KMWorld 2026 pass? Yes. Book a meeting with our team and mention it, and we can help you get a discount on your pass.

 

Contact Squirro today for a personalized demo and discover how our solutions can help you achieve unprecedented ROI.