Smarsh bets MCP and agentic AI will remake regulated comms compliance
Smarsh, the Portland-based communications-data platform that counts nine of the ten largest North American, European and Asian banks among its clients, has launched a structural overhaul of its Communications Intelligence Platform. The update introduces a Model Context Protocol (MCP) Server, a conversational AI layer called AskSmarsh AI, and a suite of purpose-built agentic tools, together designed to convert what has historically been a manual, siloed compliance function into a governed, AI-driven workflow that sits inside the tools organisations already use.
The timing is deliberate. Regulated industries, financial services above all, are now confronting a dual pressure: internal AI adoption is accelerating rapidly, yet the governance frameworks required by regulators have not kept pace. Smarsh is positioning itself at exactly that fault line, offering infrastructure that lets AI agents query live archives, surface surveillance signals, and trigger regulatory workflows without ever stepping outside a controlled permissions environment.
From archive to agentic layer
The MCP Server is the architectural pivot. By giving third-party AI tools a policy-enforced connection into Smarsh's communications intelligence data, it allows compliance teams to interrogate archives, run reports, and initiate workflows without exporting data or rebuilding governance controls elsewhere. Role-based access controls, audit logging, and chain-of-custody requirements carry over automatically to every interaction. New capture integrations with Anthropic's Claude Enterprise, Amazon Bedrock, and Google Workspace extend that governed foundation to AI surfaces that did not exist when most firms' compliance stacks were designed.
AskSmarsh AI adds a natural-language interface on top. A user can search, investigate, compile reports, and export findings in a single conversation rather than toggling between specialist point solutions, a shift the company says collapses hours of manual query-building into minutes. Three specialised agents handle the underlying work: Search, Reporting, and Export. Separately, a MySearch Agent gives individual employees self-service access to their own archived communications across email, chat, voice, mobile, and social channels without routing requests through IT or compliance teams.
The expansion of the Intelligent Agent is arguably the sharpest signal of where the market is heading. Three new non-financial misconduct models, covering hostile, abusive and discriminatory language; sexual harassment; and intent to resign, sit alongside existing financial-misconduct surveillance. That last category, intent-to-resign detection, crosses from regulatory compliance into workforce intelligence, a territory that will attract both interest and scrutiny from employment lawyers and data-protection regulators in Europe.
Cross-sector implications for capital and workforce AI
The Disrupts read-across here extends well beyond financial services compliance. The MCP standard itself, originally developed by Anthropic as a universal interface for AI agents connecting to external data sources, is quietly becoming critical infrastructure for enterprise AI deployment across sectors. Any industry that holds large, regulated archives (insurance, healthcare, government, legal services) faces the same governance bottleneck that Smarsh is solving for banks. The question for investors in this space is whether purpose-built vertical platforms, which embed governance by design, will outcompete generic AI middleware that retrofits compliance after the fact.
"AI changes the model," said Kamesh Tumsi, Chief Product Officer at Smarsh. "Customers will be able to use AskSmarsh AI to interact with their communications data conversationally, build on top of us, or bring governed Smarsh intelligence into the tools they already use."
Capital in the RegTech and agentic-AI infrastructure space is expanding, with enterprise AI governance platforms attracting growing attention from both strategic and financial investors as regulators in the US, EU and UK tighten expectations around AI-assisted decision-making in regulated environments. Smarsh does not disclose revenue or funding figures in this release, but its client concentration among global tier-one banks suggests the commercial opportunity is substantial, and the competitive response from legacy compliance vendors, as well as hyperscalers building their own governance layers, is likely to intensify through 2027.