Slides: Why your agents need decision traces, not just documents — Zach Blumenfeld, Neo4j

Source Video

Why your agents need decision traces, not just documents — Zach Blumenfeld, Neo4j

Relationship To World's Fair 2026

These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.

Related Scheduled Sessions

Extracted Slides

slide-001.jpg

OCR text:

€3 Braintrust €} WorkOS OpenAl

slide-002.jpg

OCR text:

Context Graphsi1o1 ~~

_ What They Are and How to Get Started in f

5 Minutes |

a -_ _ Sey rs a 7 oN ee

—— 7 NN Pa \

P J T

= _ ln Engineering the future of Al

oY woe om _

slide-003.jpg

OCR text:

Context Graphs Ra}

gegen Rasa cie ee

ae ae 7 om Cos Pesce

ee Sa. whe th Sy

. i RETIREE SNORE Ce Gee

r in pietegs 2 oie

ae CONTEIT GRADS CAPTURTWG Tee “Warr” q ro —. : . ae

bo oN

mae ALS cra. Hands On With Context Graphs And

inte tgeatat ee me Graphs, Graphs Everywhere-with Neo4j

° wave : Sudhir Hasbe (Ep 43) .

to nee te ae . x @:-

po D

P )

“s a | AlEngineer |

enna aed

» oe ca ad

slide-004.jpg

OCR text:

What Do Agents Need to be Accurate? ra

Knowledge Base

Information required to answer

* i questions

Ps e

po

a | AlEngineer |

oe ae ate

slide-005.jpg

OCR text:

What Do Agents Need to be Accurate? a

Knowledge Base Context Graph

Pein Information required to answer Information required to make

* i questions fe f-tol Ye ek)

Maree

ras co ee aaa ee Cea ey Cer

“Approve credit limit increase for Jessica Norns? Requesting £ 25K"

4 ; en b Engineering the future of Al

a Li 7 _ | dl

a Os

slide-006.jpg

OCR text:

Context Graph Data Model

* Person, Account, Transaction, Organization fo a

ac 7

, -

oq % ky

Decisions, Transactions, Approvals, Rejections Q

CONTEXT (The why) . -

Policies applied, Risk factors, Employee reasoning

po "

4 a | Al Engineer |

ee oil a elas

4 ee aS 4

slide-007.jpg

OCR text:

ContextGraphDemo

Financial ServicesAgent

context-graph-demo.vercel.app

DATASOURCES

BACKEND SERVICES

FRONTEND&USERINTERFACE

AIE

Ingestion/ETL

Data

FastAPI Application

Data Transfer

API Calls&

SupportTicket

+ClaudeAgent SDK

System

NexLjsUI

Application

10MCPTools

Claude Agent

User Actions

&Queies

CRM(e.g.

Salesforce)

(Embeddings)

OpenAIAPI

Chakra UIv3

Neo4jContextGraph&GDS

Internal

BusinessData

NVLGraphs

Neo

Vector

GraphData Science Algorthms

(Visualization)

Aurabs

Search

Simlaritg LouvinPageRank

(GOS)-FaRPxNNNode

PP.1dNAPKPy

Engineering the future of Al

slide-008.jpg

OCR text:

GQ Comet seen tere Bap hat ees fae AR Aree

a cc = @@ —_

acc ee 2

aw serie e@ e 6 E ~ rr

eee e oe te

not detect frm parverme e@ svaenienens

a D gintary po or

nna — _— - 7

= . er

a 7]

aan ; ,

a Wap ? Engineering the future of Al

; ie _ yt

slide-009.jpg

OCR text:

h oe A Pe fea oP catee

How the Agent Traces Decision History “ce gene

REJECT or £SCALATE ~- grounded in institutional Knowledge, not justa ced:t score “ans

we

i Query Context Graph oe

Agent reteeves Jessica's full profile accounts, transactians, employer, risk ter Heston Hin prarssenty

ran? Bee .

og ag sine ewemnce somszeas a

@ Trace Decision History iA

bd * ime

i aed Past fraud flag (Agri 2025), compliance rejections. velocny check failures

tar far Gabent_frannd_petteces|

Ej Hybrid Search Bt wee

eco beak end

Semunuc sinularnty (texatembeddings} + structural sumilunity (FastRP praph embedaings) meus | 0d ies

Reconermaditnea

+ OSCALATE FOR MAANIAL MEME

PCE ong feneog nous svete

——— ee oe

31 nodes - 30 relationships - Jessica's complete context in a single graph traversal sangitantieiconantainvalaspoed

segue ‘apesce

ry a ° °

——_

= oe Engineering the future of Al

a al

er aa . tae?

slide-010.jpg

OCR text:

rosea’ RRCAAE SEWNCE) Dixer eRe Cy.)

Context Graph Demo yi

Araceae

Cea Teena Dee Rel

Pane o

* a

* a

How GDS Powers the Agent's 10 MCP Tools mosray

ad

eran Ce ee re .

ee rare re eerye rrr

eee iris Going Forward: Industry Specific Examples

Merle ae are ore bss) Tene

fe ae re es

Pee Erneta Taree) eed peninned ae i’

Pie eee Cn re ce LATER | Government & Detenne

6 po A

‘= Al Engineer

es N PC

ws - a a vead

Ms

Peers at cane ny 4

slide-011.jpg

OCR text:

New Open Source Project: Create-Context-Graph :n

Interactive CLI ae wenn gee ey g

scaffolding tool @ euet = metinnare Cate traps MM Vuelnatnn Devnnn Treas

generates complete. See liieie E BE Ree usmeme 25 us

Pann elem Chia Qo © 3 Bes

A i wetudeet Eau HE . foes ctyert sees

a a graph applications. a \@ SET

ena od SS Bef ES ¥ sits * Spee

ara? Bo SONI Ra NK Lediimes tee

oot Sa ee ae omeeam eo = Se

for Al agents backed by “ee co. | Loe Sars

eM ee. Tee

po 7

7 ; .

gee | PA Sale lito |

a uP ae

i ie

slide-012.jpg

OCR text:

(base)zachhlumenfold:-/d

Project generated!

Generating deno data...

[1/4] Seeded 60 entities

[3/4]Generated 25documents

[2/4]Created 119relationships

[4/4]Created1edecisiontraces

Genarated:60 entities,119 relationships,25 documents,10 decision traces

Written to:/Users/zachblumenfeld/demo/my-health-apo-2/data/fixtures.json

Donel Your Healthcare context graph apo is ready.

cdmy-health-app-2

nake dockar-up

make install

Install dependencies

nake seed

Start Neo4j

AIE

makestart

Seed sanple data

start backend+frontend

Frontend:httoi//locaihost:3oe9

Backend:

http://1oc1host:80

(base)zachblumenfeld:-/demos

Engineering thefutureofAl

slide-013.jpg

OCR text:

Panes

bd ad

os

__.

= eu Al i

= pier Engineer

i ¥ q Ea Sele) a

oe 4

slide-014.jpg

OCR text:

Create-Context-Graph Key Features A

e 8 Agent Frameworks -- PydanticAl. Claude Agent SDK, OpenAl Agents SDK, LangGraph.

CrewAl, Strands, Google ADK, and Anthropic Too!s.

4 © =22 Built-in Domains -- Healthcare. finsery, real estate, and more w/ rich demo data. + You can

ra ine * create your own Custom Domains

mn . @ 12 SaaS data connectors -- GitHub. Slack, Jira, Notion, Gmail, Google Calendar. Satesforce.

Paras Linear, Google Workspace. Claude Code. Claude Al. and ChatGPT.

© Graph-native Al agents -- Cypher-powered tools for querying entities, relationships, and

decision traces. Tool calls stream in real-time with live progress indicators.

@ MCP Server for Claude Desktop -- Optionally generate an MCP server config so Claude

Desktop quenes the same knowledge graph as your web app

e Multi-turn Conversations, Streaming Chat, and Interactive Graph Visualizations

= , a Al Engineer

. a. — Sue) a

| : ae , | X

slide-015.jpg

OCR text:

Create-Context-Graph Architecture

“aoe mae 9 een)

and Cen +E es

aco =

bd a

ans

Loe C=) ==

= | =e

ee Ee} ie

See

a So i

= ; AlEngineer

‘Sy Sal ? Le

| F ca?

slide-016.jpg

OCR text:

Resources Ra)

a Create Context Neo4j Agent

a AU ar Loads abel memory

x x

ace Byer eon) Leia]

Toe Pe 2 ew =

ae Pee fee saieen

rete ey °. A # ;

(arch (al: [alee

CTE aaa a fe ee eg Pan Sates aad dC SC

eee oaTe RerT a : Tol

rea

Cue

(ass aa

‘ Pn ate oe Fa ea Tee aet Sea)

=

in CA A

ae a yt | PV Sats lag |

ma ' h oh ead

LS ag ey p: ne

Slide-Derived Subjects To Review

Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.