Chris Stapenhurst
Director, Product Management
AI Is Creating a New Surveillance Problem
Over the last year, one theme has come up consistently in conversations with compliance leaders, surveillance teams, and customers: AI adoption is accelerating faster than governance can keep pace.
That observation followed me into FINRA this year. While many AI discussions still focus on models and use cases, the conversations I had were much more operational. Firms have moved past the question of if AI belongs in the business—they’re now trying to figure out how to supervise it consistently.
AI is already showing up in drafting workflows, research, investigations, communications, search, and employee productivity tools across the organization.
The question they're wrestling with now is much harder:
How do you govern work when AI becomes part of it?
That's an important distinction because when people talk about "AI governance," they are often talking about very different things.
Some organizations focus on model governance: inventories, validation, bias testing, and monitoring for model drift. These are critical discussions, as AI systems don't remain static; models evolve, and behavior can shift over time in ways that may not be immediately obvious.
The operational challenge I hear from compliance and surveillance teams, however, sits one layer above the model itself. They are trying to understand how to govern the communications, decisions, investigations, and business records that AI helps create.
That's where the real operational challenge is emerging.
As AI becomes embedded into everyday workflows, firms are no longer just supervising employees and communications. They are increasingly supervising AI-assisted work.
That raises a new set of questions.
- How do you supervise AI-generated communications?
- How do you retain AI-assisted business records?
- How do you investigate decisions influenced by AI?
- How do you reconstruct events months later when AI was involved in the process?
- How do you demonstrate that supervisory controls operated effectively?
These aren't theoretical concerns. They're practical questions compliance teams are trying to answer today. And in many organizations, the answers are difficult because governance remains fragmented.
Communications may live in one platform. surveillance workflows in another. Investigations somewhere else. AI interactions often sit outside traditional governance processes altogether.
From a compliance perspective, that creates visibility gaps. From an operational perspective, it creates friction. And from a regulatory perspective, it creates risk.
When information is fragmented, it becomes harder to reconstruct decisions, correlate signals across workflows, explain how outcomes were reached, or demonstrate how controls operated over time. Even simple questions can become difficult to answer.
- What happened?
- Who reviewed it?
- What role did AI play?
- What controls were applied?
- Could we reproduce the outcome if asked to do so six months from now?
This is why I believe the next phase of AI adoption will be defined less by model capabilities and more by operational governance.
The challenge is no longer governing the model; it's governing the resulting AI-assisted work regardless of where it happens.
Employees may use ChatGPT, Copilot, Claude, Gemini, or the next generation of agent-based tools. Because governance cannot be rebuilt for every new platform, it needs to operate consistently across them.
That is precisely where our conversations with customers are focused today: not on governing a model, but on consistently overseeing the communications, decisions, investigations, and business records that AI helps create.
This is also where we're spending a lot of time with customers at Arctera.
The conversation is increasingly about creating a unified operational framework for governance. One that connects communications, retention, surveillance, investigations, and compliance workflows so firms can govern AI-assisted business activity with the same rigor they apply to other regulated processes.
Because ultimately, the challenge isn't governing the model. It's governing the work. And as AI becomes embedded into more business processes, that distinction will matter more than most organizations realize.
See How Arctera Helps Govern AI-Enabled Workflows
Learn how firms are using Arctera to retain, supervise, investigate, and govern AI-assisted communications and business activity across the enterprise.
Continue the Conversation
Need to Govern AI Interactions?
Ready to Operationalize Governance?
- Watch a demo of InsightAI Assistant and see how compliance teams can investigate, summarize, and act on governed communications and AI-enabled workflows faster.