Workflow Data Fabric: Connect, Contextualize, Control your Enterprise Data

[SujanDutta]

All You Need to Know About ServiceNow Workflow Data Fabric

ServiceNow Workflow Data Fabric (WDF) is the platform's unified data foundation capability. It connects to enterprise data across systems, contextualizes it with metadata and makes it available to workflows, analytics, and AI agents. Instead of copying data into ServiceNow through ETL pipelines and scheduled syncs, WDF reaches out and gets it live, at the moment when it's needed.

In this series, Karteek KVLN and Sujan Dutta break down each Workflow Data Fabric concept to help you bring enterprise data into the platform the right way.

Table of Contents

• Episode 1: Workflow Data Fabric Intro
• Episode 2: Zero Copy Connectors
• Episode 3: Fabric Tables
• Episode 4: Knowledge Graph
• Frequently Asked Questions

Episode 1: Workflow Data Fabric Intro

The Problem: Scattered Enterprise Data

AI agents are effective only when they have access to the enterprise data. On their own, agents inside ServiceNow can only work with data that already lives on the platform. That's a real limitation in any enterprise where data is spread across dozens of systems. Procurement data sits in application 1 . Analytics live in Application 2. Customer records are in Application 3. The list keeps growing with every application an organization adopts.

Traditionally, getting that external data into workflows meant building ETL jobs, writing custom integrations, syncing data on a schedule. Every sync introduces lag. Custom integrations pile up maintenance overhead. And after all this, there is high chance that the data is stale, raising governance concerns. 

The Solution: A Unified Data Foundation

Workflow Data Fabric flips that model. Instead of pulling data into ServiceNow, ServiceNow goes and gets it, live, at the moment it's needed. No batch pipelines. No scheduled syncs. No copies sitting around going stale.

WDF is ServiceNow's unified data foundation: a platform that connects to enterprise data across systems, contextualizes it with metadata and stable contracts, and makes it available to workflows, analytics, and agents.

WDF runs on three pillars: Connect, Contextualize, and Control.

The Three Pillars

1. Connect

The first job is establishing the link to wherever the data lives. WDF provides three primary mechanisms:

• Zero Copy Connectors: Real-time access to external data sources like SAP S/4HANA, Snowflake, Databricks, BigQuery, and Redshift, all without copying or moving the data. The data stays where it is. ServiceNow queries it live.
• Integration Hub: A low-code integration platform with 200+ pre-built Spokes covering common enterprise applications. Each Spoke ships with ready-to-use Actions, Flows, and Subflows for its target system. For anything not covered out of the box, the Spoke Generator builds a new Spoke from an OpenAPI spec.
• Stream Connect: Event-streaming connectivity through Apache Kafka. Direct Kafka mode connects ServiceNow to a customer's Kafka cluster as a native client, both producing and consuming messages, with no intermediary layer required.

2. Contextualize

Connecting raw data is half the job. The other half is making it understandable. WDF provides a knowledge graph that maps relationships between data assets across connected systems, along with data catalogs that give every asset business meaning. This metadata layer turns raw connections into something agents and workflows can relate to, understand what the data represents and how it relates to everything else.

3. Control

Connecting and contextualizing data without governing it creates risk. The control layer handles governance, lineage tracking, and access policies across all connected sources. You know where every piece of data came from, who can access it, and what policies apply. Data gets governed at the same time.

↑ Return to Table of Contents

Episode 2: Zero Copy Connectors

The Problem: Every Copy Starts Aging

Traditional integration follows a familiar shape. You write a job. It pulls data from the source system into ServiceNow. The data lands in a table. From that point on, you are working with a copy of data.

That copy starts aging the moment it lands. It is also a security surface you now own, and one more point of failure in the chain. Multiply that across every system you integrate with, and you are maintaining a lot of data that was true at some point in the past.

The Solution: Zero Copy Connectors

Zero Copy is the engine underneath Workflow Data Fabric, and it takes the copy out of the equation.

When a consumer queries a data fabric table, whether that consumer is an AI agent, a workflow, or a script, the Zero Copy Connector interprets the query and translates it into native SQL for the target system. That SQL goes directly to the data source. The source executes it. Results come straight back to the caller.

The data passes through ServiceNow like water through a pipe. It never sits anywhere. It never gets stored.

Pushdown Execution

For massive data sets, all the heavy compute happens at the data source. The filtering, the joining, the aggregation: that work runs where the data lives, not on ServiceNow.

That is called pushdown execution, and it is what makes Zero Copy scale to millions of records without breaking a sweat. ServiceNow sends a query and receives a result set. The source system, already built and tuned for exactly this kind of work, does the lifting.

What are the Benefits

So what are the benefits of never storing a copy of data:

• Your data is always fresh. Every query returns the state of the source as of that moment. There is no sync window to account for.
• Your security perimeter stays where it was. The data never left its source system, so the access controls and policies on that system still govern it.
• There are no ETL jobs to maintain. Nothing to schedule, monitor, or debug at two in the morning when a sync fails.
• Large data sets stay fast. Pushdown execution keeps the compute at the source, so record volume stops being a ServiceNow problem.

↑ Return to Table of Contents

In next Episode, we will understand the concept of Fabric Tables in WDF.

Episode 3: Fabric Tables

The Problem: Access the Data, which is not even stored.

Zero Copy queries data at the source instead of storing it, which raises a fair question. If the rows are not in ServiceNow, how does anything on the platform actually work with them? You have mostly build systems, where it's assumed that there is a table which can be queried by processes, whether it's scripts, flows, APIs, agents or all of it. But that's not how it works here.

The Solution: Fabric Tables

A fabric table is a ServiceNow projection of an external data object. In plain terms, it is a virtual table on the platform that carries the schema and the mapping while holding none of the data.

It looks and behaves exactly like any other ServiceNow table, but physically stores zero rows.

How One Gets Built

A data steward sets up a fabric table through a Zero Copy Connector. The connector reaches the external system through metadata collectors, which scan the schema sitting on the other side. From there the external columns get mapped onto ServiceNow field types, a primary key is set, and the table gets a name.

What Happens When You Query It

Once the table exists, it shows up in Table Explorer alongside everything else. You can write GlideRecord queries against it. You can call it from the REST Table API. You can feed it into flows and hand it to agents, basically, you use it, the way you like.

The difference sits underneath. When one of those queries runs, the Zero Copy engine wakes up, translates the request, hits the source system, and returns the results to whatever asked for them. Once that completes, the fabric table is still sitting there with zero rows.

What are the Benefits

So what are the benefits of querying through a table that holds nothing:

• Nothing to relearn: GlideRecord, the REST Table API, flows, agents leverage the same data access patterns irrespective of external data's nature.
• Nothing to rewrite: An existing script or flow can point at a fabric table, the same way it points at a local one.
• No storage footprint: The table holds the schema and the column mapping. The rows stay in the system that already owns them.
• The source stays authoritative: Every read goes back to the owning system, so there is no second version of the truth to reconcile later.

↑ Return to Table of Contents

In next Episode, we will understand the concept of Knowledge Graph in WDF.

Episode 4: Knowledge Graph

The Problem: The Data is Related, but How?

The first pillar got everything connected. Zero Copy reaches the source, fabric tables give you the data to query, and none of the data is stale. What you have at that point is still a pile of separate facts. Vendor records sit in one table, purchase orders in another, contracts and cost centers somewhere else again, and each row is perfectly accurate on its own while saying nothing at all about how it relates to the rest.

This is where the second pillar comes in, the one about contextualizing the data that you have connected.

The Solution: Knowledge Graph

A knowledge graph is where those connections are written down. It is a map of relationships across all the data you just brought in using WDF. It helps you draw the relation between each of the data points.

Semantic is the keyword here. The platform understands what the data means and how it connects to everything around it.

Ontology, Taxonomy and Metadata

Three pieces make a Knowledge graph work: Ontology defines what kinds of things exist, so the platform recognises different types of data entity. Taxonomy organizes those types into a hierarchy and places them inside your domain. Metadata describes the properties that make each one meaningful.

Put the three together and you end up with something closer to a reasoning engine than a database. You traverse it rather than query it.

How is the Graph Built in WDF

The fabric tables come first. From there you wrap them into data products for example, a vendor product, a purchase order product, a contract product, and publish those to a data catalog. Once they are published, Workflow Data Fabric draws the semantic connections between them.

These connections are typed and directional: A vendor supplies a purchase order, a contract governs a vendor, and a cost center funds an operation. Purchase orders carry line items, and those line items point at materials. None of this is code you write somewhere, it is meaning that the platform holds and understands on its own.

What are the Benefits

So what are the benefits of mapping the relationships alongside the records:

• Your AI agents can reason across the data: Following the connections, an agent can answer a query spanned across multiple tables, rather than pulling a single record and stopping there.
• The meaning lives in the platform: Relationships are held as semantics instead of being hard coded into scripts, so they hold true as the underlying data changes.
• Context travels with the data: A data product published to the catalog brings its relationships along, so whatever consumes it, understands the complete picture.
• It builds on what you already connected: The graph sits over your existing fabric tables, so nothing needs to be copied or restructured.

↑ Return to Table of Contents

Frequently Asked Questions

What is ServiceNow Workflow Data Fabric?

ServiceNow's unified data foundation. It connects to enterprise data across systems, contextualizes it with metadata and makes it available to workflows, analytics, and AI agents on the Now Platform. Instead of copying data in, ServiceNow reaches out and queries it live.

What problem does Workflow Data Fabric solve?

Enterprise data fragmentation. In most organizations, critical data is spread across dozens of systems. Traditional approaches require ETL pipelines, custom integrations, and scheduled syncs that produce stale copies. WDF eliminates that overhead by connecting to data live, where it already lives.

What are the three pillars of Workflow Data Fabric?

Connect, Contextualize, and Control. Connect establishes live links to external data through Zero Copy Connectors, Integration Hub, and Stream Connect. Contextualize adds a knowledge graph and data catalogs that give data assets business meaning and map their relationships. Control applies governance, lineage tracking, and access policies to all connected sources at the moment of connection.

How is Workflow Data Fabric different from traditional ETL?

ETL (Extract, Transform, Load) pulls data out of source systems, transforms it, and loads it into a destination on a schedule. Every sync is a snapshot that starts going stale the moment it lands. WDF skips extraction and loading entirely. ServiceNow connects to the data live and queries it on demand, so workflows and agents always see the current state. There is no copy to maintain or govern separately.

What is Stream Connect?

WDF's event-streaming connector for Apache Kafka. Direct Kafka mode connects ServiceNow directly to a customer's Kafka cluster as a native client, both producing and consuming messages on topics. No intermediary messaging layer is required, giving the customer's streaming team full control over the cluster.

What does "contextualize" mean in Workflow Data Fabric?

Raw connectivity gives you access to data, but agents and workflows also need to understand what that data means. WDF's contextualization layer provides a knowledge graph that maps relationships between data assets and data catalogs that attach business meaning to each asset. This gives agents the context to reason about the data they retrieve.

How does Workflow Data Fabric relate to AI agents?

AI agents are only as effective as the data they can access. WDF is the layer that makes enterprise data, wherever it lives, available to agents on the Now Platform. Zero Copy Connectors provide real-time access. The knowledge graph gives agents context about what the data means. The governance layer ensures agents only reach data they're authorized to see. WDF is the data infrastructure underneath enterprise AI agents.

What are Zero Copy Connectors?

Zero Copy is the engine underneath Workflow Data Fabric. When a workflow, AI agent, or script queries a data fabric table, the connector interprets that query, translates it into native SQL for the target system, and sends it directly to the source. The source executes the query and results return to the caller. The data moves through ServiceNow without being stored there.

Does Zero Copy store any data in ServiceNow?

No. The data passes through the platform like water through a pipe. Nothing lands in a table and nothing persists once the query completes.

What is pushdown execution?

Pushdown execution means the query work runs at the data source instead of on ServiceNow. Filtering, joining, and aggregation all happen inside the source system, which is already built and tuned for that kind of processing. ServiceNow sends a query and receives a result set. This is what allows Zero Copy to work against millions of records without record volume becoming a platform concern.

How fresh is the data a Zero Copy query returns?

It reflects the state of the source at the moment the query runs. There is no sync schedule or refresh window, so there is no period during which the data can drift out of date.

Does Zero Copy change my security model?

No. The data never leaves its source system, so the access controls and policies already governing it continue to apply. You avoid creating a second copy that would need its own security posture, and the source remains the single place where that data is protected.

What is a Fabric Table?

A ServiceNow projection of an external data object. It is a virtual table that looks and behaves like any other table on the platform while physically storing zero rows. The schema and column mapping live in ServiceNow; the data stays in the source system.

Does a Fabric Table store any rows?

None. It holds zero rows before a query, during a query, and after the query completes. The rows you see in a result set were read from the source and passed to the caller without landing in the table.

How does a data steward create a Fabric Table?

Through a Zero Copy Connector. The connector reaches the external system using metadata collectors, which scan the schema on that side. The steward then maps the external columns to ServiceNow field types, sets a primary key, and names the table.

Can I query a Fabric Table the same way as a normal ServiceNow table?

Yes. It appears in Table Explorer, accepts GlideRecord queries, and can be called from the REST Table API. Workflows and AI agents consume it the same way they consume any other table, so existing access patterns keep working.

What happens when a query runs against a Fabric Table?

The Zero Copy engine interprets the request, translates it for the target system, sends it to the source, and returns the results to the caller. The table itself is untouched and remains empty once the query finishes.

What is a Knowledge Graph in Workflow Data Fabric?

A map of relationships across the records you have connected. It records how your data relates, so the platform knows that a purchase order belongs to a vendor and that a contract governs that vendor, rather than treating each table as an isolated set of facts.

What does semantic mean in this context?

It means the platform understands what the data represents and how it connects to everything around it, going beyond simply holding the values. That understanding is what lets an agent traverse the data and reason over it.

What is the difference between ontology, taxonomy and metadata?

Ontology defines what kinds of things exist, such as a vendor and a contract being different types of entity. Taxonomy organizes those types into a hierarchy within your domain. Metadata describes the properties that make each one meaningful. All three are needed before the graph can hold relationships the platform understands.

What are data products and the data catalog?

Fabric tables get wrapped into data products, such as a vendor product or a purchase order product, and those products are published to a data catalog. Publishing is what allows Workflow Data Fabric to draw the semantic connections between them.

What are typed directional edges?

The connections in the graph carry both a type and a direction. A vendor supplies a purchase order, a contract governs a vendor, a cost center funds an operation. The direction and the meaning are part of the relationship, so the platform knows which way the association runs.

Why does a Knowledge Graph matter for AI agents?

Without one, an agent reads a page at a time. It can retrieve a record, but it has no way to tell what that record is connected to. With the graph in place, the agent can follow the connections across your data and reason about what it finds.

↑ Return to Table of Contents