We surveyed around 500 customer support agents in the last 3 years and we observed that around 65% of agents felt that they are unable to solve customers’ issues as quickly as they wished to. Although Conversational AI has reduced the number of tickets being raised to agents, the way the tickets are being handled hasn’t changed much in the past decade or so.
There were 3 major problems that resonated with all the agents.
1. May have to use multiple tabs/windows to perform different types of actions on different applications, leading to longer ticket handling times.
2. The need to refer to multiple documents or knowledge management systems to understand procedures to handle certain scenarios leads to inefficiencies and inaccuracies.
3. As the majority of tasks are repetitive/monotonous in nature, it usually results in a drop in motivation levels.
Dynamic AI agent workflows
We’ve described the agent assist dashboard for customer support before. But we’ve added some new features to further enhance agent productivity.
We have built this feature that has compressed an entire Standard Operating Procedure (SOP) into just a command. In the usual customer support or IT Support scenario, an agent receives a ticket that the conversational AI chatbot couldn’t resolve, and the agent is expected to follow an SOP to handle the situation based on the query/complaint of the customer.
A typical complex SOP will involve at least 5 steps before finally providing an update to the customer. It may include steps like searching for information in documents/applications, using the search results to query for information in another application, making changes in an application, updating status in an application, escalating to certain teams/team members, querying a database/application to understand the meaning of a system-generated value, etc., In most scenarios, this update may not be a resolution, but merely an escalation or initiation of an internal process.
AI Agent Workflows allow support agents to not only achieve all these actions using a single command but also allows the agents to perform actions that otherwise would have been performed by a higher level team member (L2 or L3).
Sample AI agent workflow
Based on our interactions with multiple businesses from around the world, we have come to an understanding that some businesses are not confident about allowing chatbots/voice botsto handle the conversations autonomously with customers. Since the AI Agent Workflows minimizes the risk involved in automation, the businesses can feel safe in automating parts of the customer support functions. This will result in drastic drops in OPEX of customer support teams and most importantly catapults the Customer experience to a whole new level.
The most frequently used canned response “I would request you to kindly wait for a few minutes while I retrieve the information“ can now rest in peace.
A banking customer support agent would perform the below steps or more when there is a complaint for a dispute regarding the credit card statement for the last ‘X’ number of months.
1. Query individual monthly statements.
2. Calculate the discrepancies.
3. Provide an explanation for the calculation of each statement.
4. Generate on-demand statements and compare with the previously generated statements.
5. Escalate the ticket to higher authorities if there is a discrepancy.
Sample AI agent workflow in banking
The same credit card dispute is handled by just invoking the “credit card dispute” workflow by using the command “/credit-card-dispute”.
Impact: By using this workflow, the agent will resolve this issue within 1 minute compared to the usual resolution time of 10+ minutes.
