What we heard
12 months of conversations with 50+ data team leaders.
They shared that more tables lead to significantly higher error rates.
New · A field report from PuppyGraph
What 50+ data team leaders told us about scaling agents to 1,000s of tables — and the architecture pattern that actually works in production.
01 Why we're here
12 months of conversations with 50+ data team leaders.
They shared that more tables lead to significantly higher error rates.
Ontology graphs encode rich semantics.
Text-to-SQL agents can only consult them as reference.
Most of the graph's structural power is unused.
A cybersecurity firm built a system across 1,500 tables using PuppyGraph.
Iceberg-native. Zero ETL. Customer-facing.
02 The field research
Data team leaders we've spoken with. Spanning financial services, security, networking, retail, and semiconductors.












Takeaway: every team hits the same ceiling. We've watched them climb it, then climb it again.
03 What everyone is trying
Natural language question
Context engineering
Generated query
Execution
Back to user
At small schema size: this works well.
At enterprise scale: context exceeds prompt limits + joins compound + semantic ambiguity multiplies.
04 The journey teams take
Stage 01
Start here: LLM writes SQL over Iceberg.
Workarounds teams try
Stage 02
Pivot when SQL failures compound.
Workarounds teams try
Stage 03
Double down on the model. Invest in eval infra.
Workarounds teams try
05 The diagnosis
SQL assumes the analyst already knows which joins make sense. Replace the analyst with an agent, and that knowledge vanishes — every patch is an attempt to put it back.
Reference knowledge
Lives OUTSIDE the query engine.
Enforced knowledge
Lives INSIDE the query engine.
06 The right architecture
A map is reference knowledge: useful, but ignorable. A railway is structural: the agent can only travel where the tracks go.
07 The agent-language fit
3-hop joins in SQL: ~15 lines.
Same in Cypher: 1 line.
Not a new language for the LLM.
It's a well-represented one.
Agents write Cypher.
Humans ask in English.
For agent-generated queries on graph-shaped data, Cypher is the lower-friction path. Humans don't need to learn it — they don't write it.
08 The scary failure mode
Agent attempts to retrieve student grades alongside teacher salary data — a join that violates business meaning. Production deployment becomes real when agents recover autonomously, without silent wrong answers.
SELECT s.name, s.grade, sal.salary
FROM students s
JOIN salaries sal ON s.id = sal.person_id;Error returned
No error message returned. Student suddenly has salary when they should not have.
What the agent learns
Nothing went wrong.
Returns plausible, non-empty result. Reasoning is done over incorrect data.
MATCH (s:Student)-[:HAS_SALARY]->(sal:salary)
RETURN s.name, s.grade, sal.salaryError returned
No edge 'HAS_SALARY' exists between 'Student' and 'Salary'.
What the agent learns
Salary is semantically out of scope from Student.
Error message sent back to LLM for improved query generation.
09 What it takes
The failure
Wrong joins. Silent wrong answers.
The requirement
Business rules built into the query structure. Wrong queries are structurally impossible to express, not just discouraged.
The failure
Enterprise data = massive & spread out.
The requirement
Scale and reach in one storage:
The failure
Too slow for real-time agents.
The requirement
MPP architecture and vectorized execution. Subsecond response for multi-hop traversals — agents don't wait.
10 How PuppyGraph fits in
AI Agents · Apps · Notebooks
PuppyGraphFederated Graph Query Engine · Enforced Ontology · Subsecond Performance
Iceberg-native by design. Federate across whatever else your enterprise runs on.
Query Iceberg natively. Data never leaves your storage.
Your lakehouse can be the analytical core without forcing a full migration of operational data.
11 Proof in production — 01
: Agentic IT OpsBuilt a “Glean++” agent to cut IT support costs by reducing human-in-loop ticket resolution time. Using it as the blueprint for company-wide AI overhaul.

“This work is a strong example of how we're operationalizing AI and data across the enterprise — building the foundation for more autonomous capabilities ahead.”
PuppyGraph turning point
It gave AMD's agent the context to reason, transforming a static chatbot into a self-learning knowledge engine.
Only PuppyGraph delivered the scale, speed, and live reasoning AMD needed to make enterprise AI in production.
12 Proof in production — 02
: Agentic Reasoning LayerBuilt a graph-powered digital cyber twin. Agentic workflow to dynamically compute attack vectors, blast radius, and threat paths.

“We process billions of edges on a daily basis using PuppyGraph. It's one of the products in hindsight to ask, how is this just now?”
PuppyGraph turning point
This enables AI agents to access context and answer questions like:
Try it on your tables
Zero ETL. Subsecond. Production-ready. Deploy in 10 minutes.
$ docker run -p 8081:8081 \
-p 8182:8182 -p 7687:7687 \
puppygraph/puppygraph:latestPulling puppygraph/puppygraph:latestStarting PuppyGraphWeb console listening on 0.0.0.0:8081Gremlin endpoint listening on 0.0.0.0:8182PuppyGraph is running# no pipelines · no second database · no data movement