Over the festive period, retail sales surged to $1.2 trillion globally and $282 billion in the US. It meant customer service teams were stretched thin as they strove to keep pace with customer demand.
Customers are getting increasingly impatient and expect a faster response and a better experience than ever before. A recent survey by Forethought revealed that 50% of consumers will only wait up to 9 minutes for a response from customer service. More concerning is the impact that customer disappointment delivers.
Research by PWC shows that 32% of all customers will stop doing business with a brand they love after one bad experience. Forethought’s research echoed this finding, with 23% of consumers saying they would cancel their order with an online retailer if the customer service was poor.
The state of AI in CX today
Generative AI will have a profound impact on many industries. Gartner estimates that by 2026, more than 80% of enterprises will have used generative AI APIs or deployed generative AI-enabled applications.
Many GenAI solutions focus on Retrieval-Augmented-Generation (RAG). RAG is a natural language processing technique that combines information retrieval and generative AI models. In chatbot form, it can answer questions by summarizing insights from helpdesk articles or internal resources.
Simple chatbots require CX organizations to use decision trees and create rules based on certain keywords. RAG-based chatbots can answer simple questions and resolve basic issues but cannot resolve complex issues. It creates a clunky customer experience that leaves many people demanding to speak with a human.
48% of customers believe that AI chatbots have made customer service more helpful. However, there are frustrations associated with their use, and not being able to speak with a human. The top three frustrations are:
- An automated system doesn’t provide an option to connect to a person
- Consumers have to repeatedly share the same information
- Consumers are redirected to different people and departments to solve a problem
The challenge is clear: consumers believe that GenAI can help them, but in most CX organizations, it’s not there yet. Consumers say agents who can empathize with their issues are seen as helping 95% of the time. It means that a bot who simply regurgitates FAQs isn’t going to hit the mark.
From RAGs to riches with Agentic AI
RAG-based chatbots that deliver customer experiences merely replicate what, arguably, experienced people can do better. The future of AI deployments are Agentic AI solutions.
Agentic AI is the next evolution of AI. Gartner describes Agentic AI as being able to “autonomously plan and take actions to meet user-defined goals. Agentic AI offers a virtual workforce that can offload and augment human work.”
Agentic AI promises to make customer service much more efficient and of higher quality. It can carry out RAG tasks and autonomously solve customer requests interactively. When a customer sends a service request, the Agentic AI analyses that request and identifies whether it is spam, the actual intent of the request (even if in a foreign language), and the sentiment of the customer.
Using this information, the Agentic AI can then either try to resolve the issue or request itself or pass it on to the most relevant service agent.
The communication generated by the AI can be in a tone suitable for the customer and also on brand. This can help calm the customer and ensure the language used is consistent across all engagements.
If the nature of the service request is within its remit, the Agentic AI can carry out certain tasks, including:
- Triaging the issue
- Asking questions to get a greater understanding
- Holding an interactive conversation if required
- Provide information for the customer
- Initiate actions that will help resolve an issue, including a service visit or actioning the shipment of a component or replacement. It can even initiate actions in other systems using APIs.
Keeping the human in the loop
Often, the Agentic AI can fully resolve a customer request without human interaction. If the request is beyond its scope, too complex or the customer requests a person to talk to, it will hand off to a human. It can also summarise any interactions to date and ensure that customers do not have to repeat themselves. It will even provide suggested actions for customer service agents to take next.
Agentic AI solutions initially rely on a corpus of knowledge that an organisation has. This might include a knowledge base, product information, help documents and more. As the AI resolves issues or resolves them in conjunction with a human, this knowledge can be added to the corpus. It helps the agent to answer ever more complex questions. The accuracy of this data is important. Accurate data means better responses and fewer hallucinations.
Where an AI cannot answer a question or even detect a rising frustration from a consumer, it can be handed to a more empathic human service agent. Agentic AI will make a significant difference in customer service and other business areas. Gartner predicts that by 2028, at least 15% of day-to-day work decisions will be taken autonomously through Agentic AI, up from 0% at the start of 2024. That figure might be conservative.
Agentic AI solutions do not exist in isolation. Organizations need to select a platform that provides Agentic AI and has clear governance and security. One that focuses on protecting customers and their client’s data. The platform also needs integrations at both a data and workflow level using modern APIs to enable AI agents to make a real difference within an organization’s existing operational and systems processes.
As highlighted above, data is also key. LLMs powering customer services must have access to data that is both accurate and plentiful. New additions to the knowledge base must be validated to ensure consistency, accuracy and a lack of bias. There are challenges still that can inhibit success.
Can Agentic AI deployments learn from historic failures of AI
Recent research by the RAND National Security Research Division cites the Gartner estimate that 80% of AI projects fail. Rand gave five reasons why projects have failed in the past.
- Industry stakeholders often misunderstand — or miscommunicate — what problem needs to be solved using AI.
- Many AI projects fail because the organizations lack the data to adequately train an effective AI model.
- Organizations focus more on using the latest and greatest technology than solving real problems for their intended users.
- Organizations might not have adequate infrastructure to manage their data and deploy completed AI models, which increases the likelihood of project failure.
- AI projects fail because the technology is applied to problems that are too difficult for AI to solve.
Making Agentic AI a success
Like any AI solution, Agentic AI projects can fail. However, bearing in mind the common failures, there are actions that can be taken to help ensure success for customer service deployments.
- Define a clear objective, one that will help solve a real business problem. One that is finite but scalable.
- Train AI agents on clean, qualified corporate CX data
- Select the right team, including internal consulting partners
- Communicate regularly with the team and all employees
- Select a modern cloud-based technology that understands the industry and has a ready Agentic AI solution
Why Forethought
Aptly named, Forethought was set up to do just this. The platform is fully Agentic with a patent-pending Autoflows technology that enables Agents to integrate and connect with third-party systems. The platform uses customer data while maintaining its privacy to learn and continuously learn. Integrations exist for leading solutions, including Zendesk, Salesforce, ServiceNow and Freshdesk, with many more available.
Launched in 2018, Forethought is a human-centered generative AI suite for customer support, trained on proprietary data. The company powers support for leading customer-centric organizations like Upwork, Grammarly, and iFit, and has raised $90M+ in venture capital from leading investors including NEA, Sound Ventures, and Operator Collective. Forethought has been recognized by G2 as a High Performer, Leader, and Best in Customer Support for 2024.
