Nobody questions the value of artificial intelligence in customer service operations. Customer interactions produce large volumes of high-quality, well-governed structured and unstructured data from emails, chats, transcripts, calls, and more. Unleashing AI on this data lets organizations summarize and surface key insights and predict best actions to take. The result? Better agent productivity and effectiveness that helps organizations deliver on the promise of better customer relationships, better customer retention, and increased revenue.
But this wave of AI innovation is also confusing. Customer service operations are being hit by a tsunami of AI features from every vendor, and most don't know where to start or how to deepen their journeys with AI.
More than that, core customer service products have become so over-engineered that their complexity erodes their value. Many organizations struggle to understand how to use the ever-increasing swath of features that they might not need or want. They struggle to understand how to best deploy, adopt, and onboard their (human) agents. They also struggle to understand the value of different license tiers and add-ons of all the products in the space. And AI features add a whole new layer of complexity and business decisions around their use.
Here's some guidance on what an AI adoption journey should look like:
- Level 1: Predictive AI. This is a safe way to start. Start with predictive routing, recommendations on case classification, insights on scheduling, forecasting, and trending cases. Try out next-best actions within processes and suggested knowledge articles for agents. Each of these AI use cases can be associated with increased (human) agent productivity and faster ticket resolution and can be correlated to higher customer satisfaction scores.
- Level 2: Generative AI. The safest way to try out genAI is on (human) agent-facing use cases, as agents review and can correct generated outputs. Explore summarization of content like cases, customer histories, or conversations. Use genAI to personalize service replies or draft emails that are grounded in data and knowledge. Use genAI to create drafts of knowledge base articles that must be reviewed before publishing. Quantify the ROI of each use case, looking for examples that will increase (human) agent productivity, increase their employee experience, and drive higher customer satisfaction based on more personalized responses.
- Level 3: AI Agents. Explore AI agents that are internally facing first. In a previous column I lay out a framework to explore. It recommends starting by automating simple rules-based tasks, then moving to AI agents that execute tasks with well-defined guardrails before expanding AI agents to broader workflows. ROI metrics to track include lower (human) agent utilization; increased consistency of responses, and better compliance.
- Level 4. Autonomous Agents. These AI agent frameworks are mostly aspirational at this point. They include orchestrated AI agents working together to pursue customer and business goals. AI agents learn from outputs and get better over time. ROI metrics include less (human) agent labor and ultimately true cost savings, but we are far from this reality.
Kate Leggett is a vice president and principal analyst at Forrester Research.