How AI Helps Compliance See the Risk Between the Messages

Financial Markets Compliance

November 16th, 2026

1126_How-AI-Helps-Compliance-blog_628x325

“Let’s discuss this later.”

It is the kind of message colleagues exchange every day. A meeting runs over. A question needs more thought. Someone has another call. On its own, there may be little reason for a compliance reviewer to give it a second glance.

Now imagine that message follows a conversation about a confidential deal. The exchange moves to another channel, and a trade takes place shortly afterward. The phrase still means nothing conclusive by itself. But the surrounding activity gives a reviewer a reason to look closer.

The difference is context, and it is changing what firms need from Digital Communications Governance and Archiving (DCGA). Retaining the exchange gives compliance something to examine. Connecting it to the surrounding activity helps reviewers understand why it matters. As AI makes those connections easier to uncover, firms also need a reliable way to explain the conclusions they draw.

The Conversation is Bigger than the Message

For NICE Actimize’s Paul Cottee, the progression is familiar. Early in his financial markets career, telephone calls were recorded on tapes stored near the dealing room. When a dispute arose over what someone had bought or sold, the answer was to “pull the tape”: find the recording and listen.

Those records served a clear purpose. However over time, communication records also became increasingly important to detecting and investigating misconduct. The question expanded from what someone said, to what that exchange might reveal about their conduct.

Today, the answer may be spread across email, chat, voice and mobile communications. A reviewer looking at each channel separately can miss the relationships amongst them. Even when every individual message is available, the sequence can remain difficult to see.

The phrase “let’s discuss this later” illustrates the limitation. While a keyword rule can recognize a phrase, understanding its significance requires more information about the people involved, the timing and the business activity around it. The same words can warrant very different treatment depending on those circumstances.

AI can help assemble that context by identifying participants and topics, linking related exchanges and connecting them with relevant business events. A reviewer can then examine how the conversation developed, rather than trying to infer its meaning from an isolated fragment.

The Missing Message Still Matters

There is a catch. The fuller picture is only as reliable as the information available to build it.

“Advanced AI is exciting, but it cannot fix messages that were never captured in the first place,” said Devendra Pagnis of QKS Group.

Return to the conversation about the confidential deal. If the follow-up chat never reaches the archive, an analyst may have the opening exchange and the subsequent trade but miss the discussion connecting them. More sophisticated analysis cannot reliably establish what was said in a record the firm does not have.

Building that capability takes work at every stage of DCGA. Capture and retrieval provide the foundation for monitoring, contextual analysis and AI-assisted interpretation. Each stage depends on the work beneath it.

For compliance leaders, that means going back to first-principles and examining whether all required channels are covered and whether captured records are complete and usable, alongside plans for AI. Confidence in an AI-generated explanation must begin with confidence in the evidence it uses.

Giving Reviewers Something Worth Investigating

Once that foundation is in place, context can change how teams spend their time. Adding communications channels increases the material available for review. However, generating more alerts from that material can leave analysts with longer queues and little help deciding which cases deserve attention first.

Grouping related signals and assessing them in context can help. That is, instead of reviewing disconnected alerts about the same activity, an analyst can examine a case with the relevant exchanges and supporting information brought together.

Cottee offered a practical measure of success: “The measure of AI in surveillance is not how many alerts it produces, but whether it produces better outcomes.”

That changes the purpose of automation. AI can both help reduce the effort spent sorting through low-value material and at the same time, surface connections that warrant closer investigation. Cottee suggested that skilled surveillance professionals could find themselves handling more interesting and valuable work as previously unseen risks become easier to identify.

In the opening example, the useful outcome is a reviewer seeing the connection between the message and the trade, then having enough evidence to decide whether it merits escalation. The number of alerts generated along the way says little about whether that happened.

An Explanation Someone Can Stand Behind

Bringing the pieces together also creates a responsibility to show – and be able to clearly explain - how they fit. An AI-generated summary may sound convincing, but the reviewer needs to establish whether the underlying evidence supports it.

While AI can accelerate tasks such as summarization and prioritization, judgment remains a human responsibility. A recommendation to escalate still needs someone who understands the case and can account for the decision.

That requires a trail another person can follow. For example, which communications informed the assessment? Why did their timing or relationship matter? What did the reviewer examine before deciding to act, escalate or close the case?

Cottee framed regulatory certainty around a direct question: “Can the firm prove with evidence that the program works?”

Regulatory certainty rests on demonstrating how controls operate and how decisions are reached. It means being able to reconstruct the reasoning when someone asks about it later. The context that helps an analyst recognize risk must remain available to support the decision that follows.

When AI Joins the Conversation

That reasoning becomes more complicated when AI helps create the communications under review. For example, an AI assistant may draft an employee’s response. A chatbot may answer a customer. An agent may interact with another agent while carrying out a task.

Understanding the exchange then involves more than identifying the people whose names appear on it. Firms also need to consider what role the AI played, what authority it had and who remains responsible for its actions.

Pagnis argued that governance needs to cover these interactions as they become part of everyday business. Otherwise, firms risk retaining the resulting message without enough information to explain how it came about.

Consider that routine-looking message again. A reviewer needs to understand the surrounding conversation and the related trade. If an AI assistant helped draft the message or an agent acted on it, those details become part of the story too. The value of DCGA grows with the firm’s ability to follow that story, recognize when it warrants attention and explain the action it takes.

Interested in learning more?

Watch the on-demand webinar, How AI Is Shaping the Future of Digital Communications Governance and Archiving (DCGA), with QKS Group’s Devendra Pagnis and NICE Actimize’s Paul Cottee.

    Speak to an Expert