Solve These 4 Contextual Data Problems So Agentic AI Can Find More Financial Crime

AML/KYC

The financial crime market has officially entered the Agentic AI era.

Last week, Anthropic and FIS announced a major partnership to bring Agentic AI into AML operations. Their new Financial Crimes AI Agent promises to compress investigations from hours to minutes, reduce manual work, improve SAR narratives and help investigators focus on higher-risk cases.

At the same time, Gartner issued a warning to the enterprise market:

Agentic AI without semantic context will fail.

Gartner predicts that by 2027, organizations prioritizing semantic, AI-ready data foundations will improve agentic AI accuracy by up to 80% while reducing costs by up to 60%.

These two announcements reveal something important happening in financial crime technology:

The industry is finally recognizing that AI alone is not the breakthrough.

Context is.

What Agentic AI Still Requires

The current wave of Agentic AI innovation is heavily focused on acceleration:

• Faster investigations
• AI copilots
• Workflow efficiency

Those capabilities matter.

But the effectiveness of Agentic AI in financial crime does not ultimately depend on how quickly a model can generate a response.

It depends on whether the institution has operationalized trusted investigative context underneath the AI layer.

Because financial crime investigations are fundamentally contextual problems:

• Fragmented customer records
• Disconnected transactional systems
• Unresolved entities
• Conflicting lineage
• Inconsistent data quality
• Hidden relationship structures

AI does not eliminate these challenges.

It operationalizes them.

Which means the institutions that succeed with Agentic AI will not simply deploy better models.

They will solve the foundational intelligence problems that allow AI to reason safely and effectively in regulated investigative environments.

What Has to Be True Before an AI Agent Can Find More Crime

For Agentic AI to deliver genuine investigative lift — not simply faster administrative workflows — financial institutions must first solve four foundational data and intelligence problems.

These foundations determine whether AI becomes a workflow assistant, or a trustworthy investigative capability.

1. Aggregate and Normalize Data Across Systems

Transaction monitoring, KYC, sanctions screening, customer behavior, onboarding records, adverse media, device intelligence and prior investigations must become queryable as a unified investigative substrate — not isolated operational systems.

Most banks still operate across fragmented architectures where investigative context is spread across disconnected platforms with inconsistent formats, inconsistent identifiers and inconsistent semantics.

An AI agent cannot reason effectively across fragmented context.

If the institution’s investigative data remains siloed, the AI remains siloed.

This is why contextual intelligence architecture matters far more than model selection.

2. Resolve and Deduplicate Entities

Financial crime investigations are fundamentally relationship investigations.

For example, the same beneficial owner may appear across multiple:

• Customer records
• Accounts
• Jurisdictions
• Legal entities
• multiple Identifiers

Without entity resolution and deduplication, AI agents cannot reliably identify coordinated activity or hidden investigative pathways.

Instead, connected actors appear unrelated.

This creates a critical limitation for Agentic AI systems operating on fragmented enterprise data:

the institution cannot operationalize network intelligence because the network itself has not been resolved.

Before AI can reason about suspicious behavior, the institution must first establish who and what actually exists inside the data environment.

3. Establish a Governed Source of Truth

One of the largest hidden challenges in enterprise AI is conflicting data lineage.

For example, customer attributes frequently differ across systems:

• The core banking platform shows one address
• Onboarding records show another
• Sanctions screening references a third
• External data providers introduce additional variations

An AI agent cannot safely determine which record is authoritative.

That decision requires governance.

Before AI reasoning begins, financial institutions must establish:

• Governed semantic definitions
• Deterministic lineage controls
• Evidence traceability
• Authoritative investigative standards

Otherwise, institutions risk generating fluent investigative narratives built on unresolved enterprise ambiguity.

In regulated environments, that becomes both a governance risk and a credibility risk.

4. Identify and Remediate Data Quality Gaps

Missing identifiers, unclear ownership structures, stale records, inconsistent transaction labeling, and fragmented customer histories have always created investigative risk, but those risks become far more consequential in Agentic AI environments. Large language models are exceptionally good at weaving coherent narratives from incomplete information, which means flawed data does not simply remain flawed in the background — it becomes operationalized. Weak investigative foundations can quickly turn into confident, persuasive outputs that appear credible despite gaps or inaccuracies in the underlying data.

That is why data quality can no longer be treated as a routine data management problem. In the context of AI-driven investigations and decision-making, it has become a core governance issue. And while this work is rarely glamorous, it is increasingly foundational.

But it is the work that determines whether Agentic AI becomes a productivity tool, or a genuine intelligence transformation platform.

That distinction will define the next generation of financial crime operations.

Gartner Is Right: Context Is Becoming the Core AI Control Layer

Gartner’s recent guidance on semantic and contextual AI foundations reflects a major shift happening across the enterprise market:

Context is no longer optional infrastructure.

It is becoming the operational control layer for Agentic AI.

Traditional schema-based architectures are not sufficient for complex investigative reasoning because they lack:

• Semantic understanding
• Contextual relationships
• Investigative lineage
• Governed evidence structures
• Operational meaning

For financial crime operations, this changes the architecture conversation entirely.

The Shift From Data Retrieval to Context Execution

A new operational architecture is now emerging inside leading financial institutions.

Instead of allowing AI agents to independently search fragmented enterprise systems, institutions are beginning to take a far better approach, which is to assemble investigation-ready contextual intelligence before AI engagement begins.

This changes the role of AI entirely.

Rather than asking the model to determine:

• what information matters
• which relationships are relevant
• how entities connect
• what evidence should be prioritized

the institution constructs governed investigative context first:

• linked entities
• suspicious transaction pathways
• shared identifiers
• behavioral anomalies
• organizational structures
• investigative lineage
• known typologies
• governed evidence relationships

The AI then operates inside this deterministic investigative framework.

This is the transition from: data retrieval to: context execution.

And it may become the defining architectural shift in financial crime AI over the next decade.

How DataWalk Solves the Context Problem

DataWalk approaches Agentic AI differently from traditional AML copilots and standalone LLM overlays.

Instead of relying on the model to assemble investigative context dynamically, DataWalk acts as a deterministic contextual intelligence layer across the institution’s fragmented data ecosystem.

The platform continuously connects:

• customers
• accounts
• transactions
• counterparties
• devices
• external intelligence
• organizational hierarchies
• behavioral indicators

into a governed investigative network graph.

This allows investigators — and AI systems — to operate on connected intelligence rather than disconnected records.

The impact is significant.

Deterministic Context Assembly

For example: When an alert is triggered, DataWalk automatically assembles the relevant investigative subgraph across the enterprise.

Instead of starting with fragmented data, investigators get all of the following, summarized by AI:

• connected entities
• relationship pathways
• suspicious behavioral clusters
• aggregated transaction flows
• governed evidence structures

This dramatically reduces one of the largest operational bottlenecks in financial crime investigations: manual evidence reconstruction across disconnected systems.

Preventing Agentic AI Hallucinations

DataWalk separates contextual intelligence from AI reasoning.

This architectural distinction is critical in regulated environments.

The AI model does not independently retrieve enterprise intelligence or infer unsupported relationships.

Instead, the model operates inside a governed evidence framework containing:

• validated entity relationships
• enterprise-defined semantics
• deterministic investigative context
• traceable source lineage

This grounds AI outputs in validated intelligence rather than probabilistic retrieval.

The result is higher-quality investigative narratives with significantly stronger governance controls.

Auditability and Regulatory Defensibility

Financial institutions do not simply need AI outputs.

They need explainable and defensible AI outputs.

Every generated investigative narrative must be traceable back to:

• underlying entities
• transactional evidence
• relationship structures
• investigative workflows
• governed lineage decisions

DataWalk provides full evidentiary traceability across the investigative process.

This transforms Agentic AI from a black-box assistant into a governed investigative capability aligned to enterprise compliance and audit standards.

Case Study: A Major North American Bank

A major North American bank recently used DataWalk to modernize fraud investigations at enterprise scale.

The institution faced familiar operational challenges:

• investigators spending excessive time assembling investigative context
• disconnected fraud intelligence across siloed systems
• inconsistent investigative narratives
• rising alert volumes
• pressure to scale operations without proportional headcount growth

Using DataWalk as the contextual intelligence layer, the bank automatically assembled governed, investigation-ready evidence before AI engagement.

The operational outcome included:

• dramatically faster alert-to-investigation workflows
• scalable investigative throughput
• more consistent investigative narratives
• lower operational burden
• improved explainability
• safer AI deployment aligned with governance standards
• investigators could finally focus on investigative decision-making rather than evidence reconstruction.

That is where meaningful operational leverage emerges.

The Institutions That Operationalize Context Will Define the Next Era of Financial Crime AI

The financial crime industry is entering a new phase of AI adoption.

The competitive advantage will not belong to the institutions deploying the largest models or the most copilots.

It will belong to the institutions that operationalize investigative context most effectively.

Because in financial crime investigations, speed alone is not enough.

If AI cannot operate on trusted investigative context, it cannot reliably uncover hidden financial crime activity.

It can only process fragmented information faster.

And that is the difference between accelerating workflows and transforming investigations.

FAQ

Why is contextual data more important than the specific AI model used for investigations?

While advanced models can speed up administrative tasks, their effectiveness in financial crime depends on the quality of the underlying data. Investigations involve fragmented records, hidden relationships, and inconsistent data across different systems. If an AI Agent operates on disconnected or poor-quality information, it simply processes errors faster. Establishing a unified, trusted framework of investigative context allows the AI to reason accurately and safely within a regulated environment.

What specific data problems must be resolved before deploying Agentic AI?

To ensure AI provides genuine investigative value rather than just faster paperwork, institutions need to address four foundational areas:

• Data Aggregation: Unifying records from KYC, transactions, and external media into a single searchable format.
• Entity Resolution: Identifying when the same person or business appears across different accounts or jurisdictions to uncover hidden networks.
• Governed Truth: Establishing clear rules for which data source is authoritative when systems provide conflicting information.
• Quality Remediation: Fixing missing identifiers and incomplete histories so the AI does not generate confident-sounding narratives based on faulty data.

How does a “contextual intelligence layer” help prevent AI hallucinations?

AI hallucinations often occur when a model attempts to fill in gaps or infer relationships from incomplete data. By using a deterministic layer like a graph platform, the institution separates the assembly of facts from the AI’s reasoning. The AI is restricted to operating within a framework of validated relationships and traceable evidence. This ensures that every conclusion the AI reaches is grounded in actual enterprise data rather than probabilistic guesses.

Can Agentic AI outputs meet the strict audit standards of financial regulators?

Yes, but only if the system is built for traceability. For an investigative narrative to be defensible, every claim must be linked back to specific entities, transactions, and governed lineage decisions. Moving from independent AI data retrieval to a governed framework allows institutions to provide a clear evidence trail. This transforms the AI from a “black box” into a transparent tool that meets compliance and audit requirements.

What is the practical difference between “data retrieval” and “context execution”?

In traditional setups, the AI is asked to search through systems to find what matters and determine how entities connect. This often leads to inconsistencies. In a context execution approach, the institution pre-assembles investigation-ready intelligence—such as linked entities and suspicious pathways—before the AI even engages. The AI then operates within this structured environment to summarize findings, allowing investigators to focus on final decision-making rather than manual evidence reconstruction.

What is contextual analytics?

Contextual analytics is the practice of analyzing data together with its surrounding business meaning, entities, relationships, history, permissions, provenance, and operational context. It helps organizations understand not only what happened, but what it is connected to, why it matters, and what action should follow.

What is data contextualization?

Data contextualization is the process of turning raw or fragmented data into meaningful, connected, reusable context. It connects records to real-world entities, relationships, definitions, permissions, provenance, and history so teams can work from a shared understanding rather than isolated tables, files, or extracts.

What is Persistent Context?

Persistent Context is reusable, governed context that survives beyond a single project. It allows new sources, workflows, investigations, applications, models, and AI initiatives to extend an existing intelligence model instead of rebuilding context from scratch.

How does DataWalk create Persistent Context?

DataWalk maps data to a flexible ontology and maintains entities, relationships, permissions, provenance, and analytical structure in one extensible intelligence model. New sources and use cases extend this model rather than creating disconnected context in separate marts, graphs, or knowledge bases.

Is DataWalk a knowledge graph?

DataWalk uses ontology and graph concepts, but it is broader than a standalone knowledge graph. It operationalizes connected context across entity resolution, search, graph analytics, investigation, scoring, workflows, applications, APIs, MCP, and AI access on one governed context layer.

Does DataWalk replace a data lake or warehouse?

No. DataWalk does not replace existing data lakes, warehouses, or source systems. It connects data from those environments and turns it into reusable context for analytics, investigations, workflows, applications, decisioning, and AI agents.

Why does Persistent Context matter for AI?

AI needs more than clean data. It needs governed context such as resolved entities, explicit relationships, permissions, provenance, definitions, and controlled analytical functions. Without this, AI systems often rely on isolated files or extracts, which leads to weak explainability, poor governance, and inconsistent answers. DataWalk provides a persistent context layer that AI agents and models can consume through APIs, MCP, and analytical functions.