Whether it’s helping a patient understand a medical bill or providing risk-based investment information, businesses depend on customer service agents for customer satisfaction and loyalty. Courteous human agents can answer a wide range of questions quickly, which is what customers want. But when business customers have complex issues, those reps often struggle to access and process the right information fast enough to resolve the issue.
Enter AI agents. Chatbots and similar indefatigable assistants have the capacity to answer any number of straightforward requests 24/7. Their nonstop activity frees up human customer service agents to give more personal or in-depth attention when needed.
The latest evolution in virtual assistants, agentic AI, takes the chatbot concept several steps further. Equipped with enhanced reasoning capabilities, these semi-autonomous tools work either alongside human agents or directly with customers to solve complex, multi-step problems.
Using generative AI (GenAI) and large language models (LLMs), agentic AI can both think and act with much more autonomy. A simple example involves a customer email about a delayed shipment. Typically, a human customer service agent would check the order, check inventory and possibly forward the email to another representative. The customer might get an answer or resolution days later. Agentic AI would “read” the email, check inventory and other systems, flag the issue, notify logistics, initiate reshipment if needed, and reply to the customer with an apology and confirmation of the reshipment, all in a fraction of the time a human-only process would take. And, agentic AI would complete all these tasks without human intervention.
To operate at that kind of speed and scale, agentic AI must analyze volumes of documentation in near-real time. To do so requires not only smart AI-based reasoning models, but also a powerful, secure infrastructure designed for that kind of data processing load. Without the right infrastructure, an organization could also face unexpectedly high costs from its cloud service provider when they scale AI.
Deploying AI at scale more cost-effectively requires what’s known as serverless inference, where cloud-provider-managed infrastructure resources are spun up and scaled automatically based on the AI workload and use-case requirements. Because the resources are in use only when needed, the organization can avoid over-provisioning massive GPU and CPU resources for their AI agents.
Use cases: Agentic AI in action
Agentic AI is emerging and receiving strong consideration. A 2024 survey from Forum Ventures showed that 48% of senior IT leaders surveyed are beginning to adopt AI agents, with an additional 33% actively exploring solutions.
Here are three other use cases that show the potential of agentic AI.
Health insurance companies: Agentic AI can process thousands of insurance claims at once. Using advanced reasoning, the tool can be trained to autonomously triage, validate and resolve insurance claims based on policy terms, medical codes and patient history.
The AI agent could also, using HIPAA-compliant channels, request missing information from providers, cross-reference diagnostic codes and approve or deny claims with minimal human supervision. When it catches an anomaly, it could flag the item for human review. The agent would save the company thousands of hours annually in manual processing time, while also reducing errors, oversight and fraud.
Financial investment firms: The average financial services firm may track hundreds of equities, sectors and macroeconomic indicators across global markets. While monitoring all that activity, analysts rarely have time to review the earnings reports and other news that could affect prices.
Agentic AI can act as a 24/7 market intelligence assistant, analyzing incoming data to help financial advisors respond to volatility in real time. Because the agent doesn’t need to sleep, it can support reps in any time zone.
On the factory floor: A technician on the production line encounters a malfunctioning piece of equipment on the factory floor. Every minute of downtime costs thousands in lost output, leading to shipment delays. The technician would need to search through a 200-page manual to fix the problem, losing valuable time. Shipment delays could lead to either irritated customers or lost business as well as reputational damage.
Agentic AI integrated with factory sensors, maintenance logs and equipment documentation could diagnose the issue in minutes and guide the technician through repairs. In the process, it would recommend troubleshooting steps and replacement parts for rapid resolution. As it learns over time, the agent could even flag known precursors to the issue before it causes a malfunction again.
How to build an agentic AI infrastructure
Agentic AI requires a more complex infrastructure than basic chatbots or automation. Serverless inference, combined with retrieval-augmented generation (RAG) where the data resides, provides a solid foundation while protecting sensitive data. RAG combines a large language model (LLM) with a retrieval system that accesses a specific knowledge base, which allows the model to work with up-to-date and contextually relevant information. Appropriate security measures, such as access control and data encryption, protect sensitive data from leakage.
Vultr provides high-performance compute, storage and GPU resources in the cloud for scalable and efficient data handling. It is well positioned for a serverless inference approach, offering capacity and security at a cost that’s far less than the top three cloud providers.
For agentic AI, one recommended approach involves combining the Vultr cloud infrastructure with NVIDIA and NetApp. The NVIDIA AI Blueprint for RAG, combined with NVIDIA NeMo Retriever built with NVIDIA Inferencing Microservices (NIM), can create a custom AI workflow. These microservices help developers build AI applications that find, contextualize and extract information from text, tables and images, including both photos and videos. NetApp meanwhile enhances data management with optimized data movement and access while delivering compliance-assured, secure storage.
Agentic AI is the next wave
Agentic AI is the next step in the AI transformation that’s upending almost every industry. It also has the potential to dramatically improve customer service by automating and even completing tasks with little to no human intervention.
With a serverless inference approach, an organization’s agentic AI will be ready to both reason through and act on complex customer questions and concerns. Meanwhile, organizations can benefit from faster response times, more efficient customer service operations and a blend of personable human interaction and dependable, tireless AI agents.