Smarsh deployed two AI agents built on Salesforce's Agentforce platform that are reshaping customer and internal support workflows. Its customer-facing agent, Archie, resolved 72% of interactions without human help in Q2 2026, while an internal agent called Emmy saved support representatives an average of 7.5 hours per complex case. For customer support teams in regulated industries, the results offer a measurable blueprint for adding AI without sacrificing compliance or control.
Archie moves from pilot to production with measurable results
Launched last year, Archie is an AI-powered customer support agent designed to handle routine inquiries for the communications data and intelligence company. Out of 405 interactions in the second quarter of 2026, more than seven in 10 customer sessions were resolved without escalating to a human agent. Archie also earned a 2.6 out of 3 resolution confidence score, a metric Smarsh developed to measure whether AI responses fully addressed customer questions without frustration or follow-up within 48 hours.
"A year ago, we set out to use AI in a way that would create tangible value for our customers and our teams," said Rohit Khanna, Chief Customer Officer at Smarsh. "The most important result isn't simply that Archie can resolve routine needs faster. It's that our customers can trust us to get a consistent experience while our people have more time to focus on the complex issues where their knowledge and judgment matter most."
Emmy targets internal workflows and cuts case resolution time
Smarsh extended the same approach inward with Emmy, a second Agentforce agent launched in late March 2026. Emmy gives support representatives instant account snapshots - ownership, status, case history - plus AI-assisted resolution guidance at case closure. By the end of June, 120 of 185 representatives had used Emmy, a 65% adoption rate that exceeded the company's 50% target.
Complex Smarsh support cases often involve system-log analysis and engineering escalations. Representatives using Emmy saved an average of 7.5 hours per case compared with similar cases handled before the agent was introduced, clearing the five-hour target the company had set. Across all cases, 31% were resolved with AI assistance, above the 20% initial goal.
"Smarsh is a standout example of how Agentforce is helping enterprises scale trusted AI in the way a regulated industry like financial services demands," said Greg Beltzer, Chief Customer Officer, SVP-Agentforce at Salesforce. "The results from Archie and the early success metrics from Emmy show how Agentforce helps organizations improve service and increase productivity, while meeting the governance financial institutions need as AI becomes more deeply embedded in their operations."
A repeatable model for regulated support teams
Smarsh serves many of the world's largest banks, insurers, and government agencies - organizations that cannot compromise on compliance rigor even as they adopt automation. The company has also received a Notice of Allowance from the U.S. Patent and Trademark Office for the Archie trademark, signaling intent to protect the AI brand as it builds out additional capabilities.
For support leaders looking to apply similar tools, the metrics from Archie and Emmy provide concrete benchmarks: deflection rate, resolution confidence, adoption percentage, and hours saved per case. Teams exploring AI for Customer Support Courses can use these figures to set internal targets and evaluate vendor claims. Smarsh will share implementation lessons at Dreamforce on September 16, where Phil Dean, Vice President of Global Customer Support, joins a service leader panel on scaling AI-powered service.
Why this matters for customer support professionals
Archie and Emmy are not experiments - they are in production with adoption rates, time savings, and resolution data that any support leader can measure against. The 72% deflection rate means fewer repetitive tickets land on agent queues. The 7.5 hours saved per complex case frees senior representatives for work that demands judgment. For teams operating under regulatory scrutiny, Smarsh's approach shows that agentic AI can hit efficiency targets without loosening governance controls. The next step is auditing your own case types to identify which ones match the profile Archie handles - routine, high-volume, answerable from existing knowledge - and piloting an agent there first.
Your membership also unlocks:
