The AI communications data and intelligence company, which provides its own AI agents for compliance and surveillance to customers in highly regulated industries like financial services, is leveraging Salesforce Agentforce agents for a variety of tasks.
Rohit Khanna, chief customer officer, Smarsh
Fintech Smarsh announced today it has deployed Salesforce’s Agentforce platform to power AI agents for customer support.
Smarsh is no stranger to agentic AI. The company specializes in providing compliance and surveillance technology, including proprietary AI agents, to customers in industries such as financial services and government. Its more than 6,200 customers include 18 of the top 20 global banks.
“Our focus has been on compliance-related intelligent agents,” says chief customer officer Rohit Khanna.
Khanna is responsible for all post-sales activities, including technical support, professional services, managed services, and implementation of data engineering, data systems, and data migrations. Plus, he oversees Smarsh University, which trains customers on Smarsh solutions.
Existing strengths improved
Smarsh captures the electronic communications of all regulated users, mostly in financial services, and stores that captured data according to the regulations that apply. It also surveils captured data for misconduct like insider training or other forms of market manipulation.
“When they’re on their WhatsApp, email, SMS, LinkedIn, and Facebook, we capture it all so we know what they’re telling their customers, and what they’re telling each other,” Khanna says.
While his global team of 400, across nine centers of excellence, is very experienced with autonomous agents for compliance, Salesforce’s expertise with support-related agents made Agentforce an easy choice for Smarsh’s customer service needs.
“Because Salesforce is our CRM and where our master data is, and because Salesforce provides us our ITSM through Service Cloud, it just made sense for us to go acquire agents from them,” he says.
Putting use cases to use
Last year, Khanna says the Smarsh board started inquiring about how AI could increase the efficiency of the organization. After some research, the team identified a portfolio of initial use cases in Q4. It acquired the Agentforce platform in Q1 2025 and started implementing customer cases in Q2.
“The number one use case to really get used to it was the creation of automatic knowledge base articles,” Khanna says.
Smarsh began developing its knowledge base several years ago, and while that endeavor utilized another AI tool, it still meant support engineers around the world were charged with creating thousands of knowledge base articles. As of Q2, though, an agent creates new knowledge base articles autonomously.
The second use case was a common one: summarizing meetings. The third, which Smarsh is formally announcing today, is the replacement of its customer-facing chatbot on Smarsh Central with an Agentforce Service Agent it calls Archie, complete with an animated avatar that gives a face to the agent.
The chatbot used a Q&A format but for the past several months, Archie has been writing responses to customer cases. Smarsh has kept a human in the loop for these responses, with support reps responsible for looking at the customer cases, and deciding whether to approve Archie’s responses. At the beginning, Khanna had 20 support reps on the task, but now he’s reduced it to five. If Archie can’t adequately answer a customer’s question within three prompts, the system automatically routes the customer to a live agent.
“It’s getting better with time,” he says. “I always tell my team, don’t blame the messenger, because the message is coming from your knowledge base. If our data is wrong or we have gaps in our knowledge base, you can’t blame the messenger. What we’ve seen is it’s not the tech that’s the problem, it’s the data.”
Khanna’s team is now in the process of rolling out a fourth agent for billing, which will be followed by an agent to help customers with user access control.
Something in the numbers
Khanna says Smarsh expects to see a 20% increase in customer self-service success rates, 25% faster issue resolution compared with the previous traditional self-service search and browse methods, and a 30% increase in service representative productivity.
That last metric, he says, directly answered the board’s concerns over efficiency.
“The Smarsh Global Services team grew 30% this year but we actually didn’t hire any level one support reps at all,” Khanna says.
In fact, he doesn’t expect to hire any level one support reps next year either, as the goal is now to allow agents to handle that level of support completely.
“We didn’t let people go,” Khanna stresses. “We just halted the growth. Now we’re challenging and pushing our team members to sharpen their skills and go to level two or level three, what we call business critical support.”
By the end of next year, Khanna is aiming to have zero level one reps in the business, with all existing team members in that role skilled up to level two or three.
“Everything will be replaced,” he says. “All level one at my COEs, whether in the Philippines, India, Belfast, Costa Rica or US, will be replaced by service agents, and level two and level three of business-critical support will be our manual support.”
Khanna adds that Smarsh recently changed their titles from customer representatives to technical representatives or technical engineers.
