Most businesses use chatbots to provide resources for quickly solving simple issues. This is an example of businesses meeting customer expectations. Salesforce’s State of the Connected Customer estimates that 61% of customers want to use self-service to solve basic problems. But according to the same study, 68% of customers wouldn’t use a chatbot again if they had a bad experience.
So far, developments with AI support chatbots have been more sizzle than steak, especially when compared to large language models (LLMs) like Claude and ChatGPT. As LLMs become more widely used, disappointing interactions with AI support chatbots can further complicate already frustrating situations.
Common issues with AI agents include misinterpreting inquiries and providing inaccurate responses. Traditional chatbots can only handle pre-programmed queries. When they’re stumped, customers often have to repeat themselves to human agents, leading to frustration about time wasted. It shouldn’t come as a surprise that Fast Company found that 81% of customers prefer speaking to a live agent, even if it means waiting. Currently, chatbots aren’t meeting expectations for quick, personalized service. The saying "you never get a second chance to make a first impression" doesn’t only apply to humans anymore.
Salesforce has responded by rolling out the next generation of chatbots. One of the first autonomous agents introduced is the Agentforce Service Agent. These bots interact with LLMs, analyzing the context of the customer’s message. Generative AI built on the Einstein Trust Layer uses your business’s data to create messages. Building on the Einstein Trust Layer ensures that you can safely train bots using your data and trust that data remains secure. These agents are trained on the language and support processes used today. Messages generated by the bots align with your brand, tone, and service guidelines. But how do you determine what tasks human agents do vs. those done by Agentforce Service Agents?
To answer that question, we need to take a step back and consider the spectrum of support inquiries. Most support inquiries fit into one of three tiers: Action-Based, Order-Based, and Knowledge-Centered. Agentforce Service Agents can help agents in each tier perform their jobs more efficiently. When human agents handle a case, the resolution cost increases with the time spent on it. Longer case handle times mean service profitability is affected.
Action-Based cases are simple interactions between customers and agents. These exchanges will feel similar to a “meeting could have been an email.” They are quick or repeatable interactions that are often resolved on first touch. A simple password reset is a quick example of an Action-Based case. Given the choice, many customers prefer self-service options to resolve these issues. A successful case deflection strategy can result in 40%-70% of these interactions never reaching an agent.
Companies with higher percentages of case volume falling into the Action-Based category will yield the best value from Agentforce’s Service Agent technology. If human agents work fewer of these cases, they can take on a higher volume of complicated interactions.
Order-Based cases are order entry, light troubleshooting, or answering basic inquiries. These case types will see low-value tasks replaced by Agentforce Service Agents. The remaining tasks will be those of higher value to the customer or company. The customer experience is as valuable as the product itself, so some human interaction may be necessary to provide the best possible experience. Deflecting 25-40% of these cases is considered good for these types of cases. For service teams, this means they can offload tedious tasks that slow productivity, freeing human agents to focus on more meaningful, human-centric tasks.
Finally, Knowledge-Centered tasks require highly trained, experienced, human reps. These interactions are specialized and nearly impossible to replicate, making them a core focus for your service team. Agentforce Service Agents start these conversations and hand off cases to a human counterpart based on parameters set by the company.
Agentforce Service Agents can interact with customers through Salesforce’s Messaging for In-App and Web (MIAW) chat, voice, SMS, and Slack. As service quality improves, Agentforce Service Agents continuously train on this progress, further improving performance. With AI agents seamlessly handling routine tasks and enhancing human interactions, businesses can finally deliver the personalized, efficient service their customers expect—at the speed of tomorrow.
Is your org AI-ready? Stay tuned for our upcoming blog, where we'll cover the foundations of building effective service processes that Agentforce Service Agents are trained on. Learn how to prepare your Salesforce org to unlock the full potential of Agentforce.
