Smarsh EDGE: Governance becomes foundation for communications intelligence

A study recently released by Smarsh Inc. found that only 26% of enterprises say their governance frameworks are keeping pace with AI deployment, even as AI in financial services moves beyond pilots, according to the report. The company and its partners described compliance and observability as prerequisites for using AI across email, messaging, collaboration and AI-prompt traffic, including agents communicating with other agents.

Goutam Nadella, Smarsh's chief strategy officer, said the company is helping customers extract insights from communications that had been too unstructured to use. A customer relationship management system holds only what someone recorded about a meeting, while AI can classify a year of conversations across email, WhatsApp, Teams and Zoom and suggest how to prepare for the next one, he said. Smarsh was a launch partner for compliance application programming interfaces from Anthropic PBC and OpenAI Group PBC, capturing prompts so risky outputs can be traced to their sources.

Nadella said governance must be foundational rather than episodic. He pointed to billions of dollars in regulatory fines imposed on financial institutions for failing to preserve business communications conducted through WhatsApp and other messaging platforms. The report also said customers are building their own agents outside compliance platforms and accessing data through Model Context Protocol servers; regulations remain technology-agnostic, and financial institutions remain liable for their agents' actions.

Rohit Khanna, Smarsh's chief customer officer, said supervisors who once ran complex Boolean searches to find market manipulation can now ask an agent questions in plain language and receive flagged conversations with risk scores. A context layer that maps identity hierarchies, for which Smarsh has filed a patent, helps prevent the AI from generating false positives, he said. Smarsh also builds compliance language models that are cheaper to run than large language models, including a multilingual model that tracks speakers switching between languages. Its S-Combinator innovation group of 11 volunteers built a model risk management agent now in limited availability.

Kamesh Tumsi, Smarsh's chief product officer, said organizations can combine Smarsh's agents with their own models and agents while the platform provides governance, data access controls and explainability underneath. Smarsh has launched four or five agents in six months alongside its new MCP server, according to Tumsi. Because 70% to 80% of compliance spending still goes toward labor, he said the potential return from AI is significant. He described four levels of autonomy: assisting, recommending, executing under guardrails and fully autonomous, with each process matched to a level by weighing whether it is reversible, how complex it is and how much observability it requires.

On a panel with Arym Diamond, head of data and AI go-to-market sales at Amazon Web Services Inc., and Peter Sankauskas, portfolio chief technology officer of Caylent Inc., Nadella said customer anxiety about AI a year ago has turned into excitement now that the infrastructure exists. Through a multiyear strategic collaboration agreement with AWS, Smarsh is building governance and security in at the infrastructure layer, because customers now want to know who or what is accessing their data, according to the report. Smarsh has shifted from a system of record to a system of engagement, the report said.

Gemma Allen, host at theCUBE, SiliconANGLE Media's livestreaming studio, said the move from system of record to system of engagement to system of intent is "really won or lost in one battleground, and that is data." She added that data is "a huge moat for companies in 2026."