Regulated industries like finance and healthcare demand support systems that balance speed with ironclad security. Smarsh, a provider of cloud-native communications archiving for these sectors, faced a familiar challenge: customers drowning in fragmented documentation and compliance hurdles. The solution wasn’t just another chatbot—it was an AI-powered ‘front door’ that simplifies access while preserving the rigor of highly controlled environments.

The result is Archie, an AI agent trained on Smarsh’s proprietary knowledge base. Unlike traditional self-service portals that force users to navigate labyrinthine menus, Archie lets customers describe their needs in plain language. The system then routes them directly to the right resources, slashing resolution times by 25% while boosting self-service adoption to 59%.

Key specs of the deployment

  • Platform: Salesforce Agentforce 360 (orchestration layer for agentic workflows)
  • Data foundation: Five years of pre-cleaned, annotated, and anonymized documentation
  • Compliance integration: Salesforce Trust Layer for audit-ready model risk management (MRM) approvals
  • Performance: 20% increase in self-service success rates; 30% productivity gain for support teams
  • Adoption trigger: Natural language input with zero-code setup for regulated users

The architecture hinges on two principles: unified data and controlled execution. Smarsh avoided the pitfall of many AI pilots—where projects stall at the ‘last mile’—by embedding Archie into Salesforce’s existing ecosystem. This ensures the AI can execute workflows across systems without compromising compliance. For example, when a customer asks about archiving a financial communication, Archie doesn’t just point to a PDF; it triggers the correct compliance workflow in the background.

Data hygiene was non-negotiable. While many companies rush to deploy generative AI only to hit roadblocks with messy datasets, Smarsh spent years rationalizing its documentation. Every guide, policy, and technical note was annotated, anonymized, and locked down—preparing for the day AI would need to parse it. ‘We started strong because our data was already clean,’ says Rohit Khanna, Smarsh’s chief customer officer. ‘Today, we’re in production with an agent that regulators can audit.’

Regulatory scrutiny remains the biggest hurdle. Financial institutions and banks demand model risk management (MRM) approvals before adopting AI tools, forcing Smarsh to prove Archie’s outputs are traceable and secure. Salesforce’s Trust Layer provided the necessary framework to explain to auditors how data flows through the system—without exposing it to external LLMs. ‘They ask, Tell me the LLM isn’t leaking our data,’ Khanna notes. ‘Salesforce’s documentation helped us answer that.’

Adoption wasn’t automatic. Early confusion over the text-based interface (some users expected a traditional chatbot) required a shift to personalized onboarding. Once customers realized they could ask questions like ‘How do I archive an email with a client?’ in natural language, engagement surged. The lesson: AI in regulated spaces needs both technical precision and human-centric design. With Archie now live across Smarsh’s core products, the company aims to replicate the 59% adoption rate enterprise-wide—proving even the most constrained industries can unlock AI’s potential.