Can human investigators keep pace with AI-powered deception?
For decades, intelligence work has followed a remarkably consistent pattern.
Whether investigating financial fraud, verifying identities or assessing geopolitical risk, analysts have traditionally done much the same thing: formulate a hypothesis, search for information, compare sources and gradually assemble enough evidence to reach a conclusion.
Artificial intelligence is beginning to change that equation on both sides.
The same technology helping organisations automate workflows and accelerate decision-making is also making deception significantly easier to scale. Synthetic identities can now be assembled from fragments of genuine personal information. Fraud networks can generate convincing documentation in seconds. Coordinated influence campaigns can produce multilingual content at a speed that would previously have required large teams.
If deception increasingly operates at machine speed, investigations cannot rely primarily on human bandwidth, argues Jen Snell, chief marketing officer at Babel Street, in this exclusive interview.
In 2026, Babel Street integrated its Data Dominance repository with Agentic AI via tools like Insights Investigator. Instead of requiring analysts to manually dig through millions of data points, autonomous AI agents are assigned complex research plans. Fueled by Data Dominance, the foundational capability and primary data infrastructure of the Babel Street Platform, these agents sift through vast digital noise to map illicit networks and uncover hidden threat vectors at machine speed, always returning structured findings with clear citations and full source provenance.
Babel Street's new Agentic AI strategy is a response to what chief executive Benji Hutchinson called "the AI-on-AI era"—a world in which artificial intelligence is increasingly used both to create sophisticated threats and to detect them. Rather than positioning AI as another productivity assistant, the company argues that intelligence workflows themselves need to evolve. Its subsequent launch of Insights Investigator marked the first major product built around that vision.
"Risk intelligence has become more urgent as threats have grown more sophisticated and difficult to detect," said Snell.
"Organizations are making high-stakes decisions about suppliers, identities, partners, and emerging threats based on intelligence that is often incomplete, outdated, built on dangerously narrow data, or too slow for timely action."
According to Snell, the problem is no longer simply finding information. It is processing enough information quickly enough to expose patterns that increasingly unfold across thousands of interconnected data points.
That distinction may sound subtle, but it represents a fundamental change in how intelligence systems are being designed.
Traditional platforms function rather like highly sophisticated search engines. Analysts formulate questions, retrieve relevant information and then perform much of the analytical work themselves. Agentic systems reverse that relationship. The Babel Street platform, specifically, has evolved from a platform that analysts search to a platform that analysts command, as Snell explained. Instead of constructing complex queries, analysts define an investigative objective. The AI then traverses large datasets, extracts entities, resolves identities, identifies anomalies and assembles findings before presenting its conclusions alongside the underlying evidence and reasoning.
The new economics of deception
Babel Street's strategy is built on the assumption that AI is changing not just how organisations work, but how criminals operate.
"The only way to counter AI-powered threats is with AI-powered risk intelligence," Snell argued.
Her reasoning extends beyond generative AI itself: the company points to four areas where conventional intelligence workflows are coming under increasing pressure.
The first is identity verification.
Fraudsters are no longer limited to stealing identities. Increasingly, they create synthetic ones by combining genuine personal information with fabricated details that appear plausible enough to pass conventional verification checks.
A static identity check may confirm that a person appears to exist. It says far less about whether that identity has been artificially constructed or whether seemingly unrelated applications are linked through shared addresses, telephone numbers or other hidden relationships.
The second challenge is organisational fragmentation.
Fraud rarely confines itself to a single institution or government programme. Yet information often remains trapped inside organisational silos, making it difficult to recognise coordinated activity that spans multiple systems.
The third concerns language.
Much commercial threat intelligence continues to focus primarily on English-language sources, despite many emerging risks first appearing elsewhere. Babel Street says its platform analyses information across more than 200 languages in an effort to reduce those blind spots.
Finally, there is network analysis.
Many organised fraud schemes only become visible once investigators stop examining individual identities and begin analysing relationships between people, organisations, locations and digital activity. Graph-based analysis has existed for years, but AI increasingly makes it possible to perform that work continuously and at far greater scale.
Taken together, these four justify the use of agentic AI: datasets have grown too large, too dynamic and too interconnected for humans on their own to discover meaningful relationships in.
Human judgement doesn't disappear
One of the recurring concerns surrounding agentic AI is whether giving software greater autonomy inevitably means giving up human control.
Babel Street argues that the opposite should happen.
Rather than replacing analysts, the company positions AI as a specialist colleague responsible for the most labour-intensive parts of an investigation—collecting information, following relationships across datasets and surfacing evidence—while humans remain responsible for defining objectives, testing conclusions and making decisions.
"The AI-as-a-Worker model is built on a foundational principle: human judgment remains essential," Snell said. "This is especially true in regulated, legally consequential, or national security environments."
Instead of spending hours manually gathering information, analysts are expected to devote more of their time to interpretation, critical thinking and oversight.
That distinction is becoming increasingly important as AI systems move into domains where mistakes can have significant consequences.
An incorrect recommendation from a chatbot may be inconvenient.
An incorrect investigative conclusion affecting financial services, law enforcement or national security could be considerably more serious.
Babel Street says this is why its platform exposes the reasoning behind every investigation rather than treating the AI as a black box. Analysts can review the research plan, inspect the query logic and examine the evidence supporting each conclusion before making decisions. Audit trails record every stage of the process so findings can be reproduced and defended if challenged.
That emphasis on explainability reflects a broader trend across enterprise AI.
As organisations begin deploying increasingly autonomous systems, transparency is becoming almost as important as accuracy. In regulated industries, organisations need to understand not only what an AI concludes, but also how it arrived there.
The more responsibility AI assumes, the more important that distinction becomes.
It's all about the context
Snell argues that many intelligence platforms create a false sense of confidence because they draw conclusions from relatively limited information. Richer contextual datasets, by contrast, allow AI systems to identify anomalies, recognise hidden relationships and build a more complete picture of potential risk.
"Our Chief AI Officer John Larson has underscored this, saying, 'AI falls short today not because it lacks sophistication, but because it is fueled by data that lacks context,'" she said.
One example involves synthetic identities.
Someone claiming a twenty-year residential history but leaving virtually no digital footprint—no local activity, no online presence and no other supporting signals—may warrant further investigation even if traditional identity checks indicate that the individual exists. Context, rather than isolated data points, becomes the differentiator.
This is why Babel Street Data Dominance™ comes in: as the foundational capability of the Babel Street Platform, it is designed to transform massive volumes of chaotic, public open-source information into structured, contextual intelligence. Babel Street says the tech can process and interpret information across more than 200 languages, for example, going beyond basic translation to preserve local nuance, slang, and coded language. Data Dominance also ensures that all data in the estate is compliant, legally cleared, and audit-ready to safeguard high-stakes government and enterprise workflows.
From answering questions to conducting investigations
Babel Street's roadmap extends beyond today's product.
The company says it is developing agent-to-agent interoperability that would allow external AI systems to interact directly with its platform to enrich investigations, while future versions of Insights Investigator are expected to incorporate multimodal capabilities that combine textual and visual information into a single investigative workflow.
Whether that vision becomes commonplace remains uncertain.
The broader direction, however, increasingly resembles developments taking place elsewhere in enterprise AI.
Across software engineering, industrial operations, customer support and knowledge work, AI is gradually evolving from answering individual questions towards executing increasingly substantial parts of complex workflows under human supervision.
Risk intelligence appears to be following the same trajectory.
At the same time, that evolution raises important questions.
Relationship mapping and autonomous investigations are only as reliable as the data they analyse. Incomplete information, incorrect links between individuals or flawed assumptions could all produce misleading conclusions, particularly in high-stakes environments. As organisations place greater trust in AI-generated investigations, governance, transparency and human accountability become just as important as speed.
The challenge, in other words, is no longer simply building AI that can investigate.
It is building AI whose investigations people can trust.
Early results suggest the approach has practical potential. Babel Street points to work with the Center for Intelligence Research, Analysis and Training (CIRAT) at Mercyhurst University, where the organisation says analysts reduced the effort required to produce intelligence outputs by around 50% while saving several hours per analyst each week.
What already seems clear is that investigations built entirely around human bandwidth are being challenged by a world in which deception itself increasingly operates at machine speed.
Further reading on MoveTheNeedle.news:
Yubico targets the next phase of AI security through OpenAI partnership