All over the world, organizations are finding ways to incorporate AI into everyday work, from helping employees find information faster to automating routine tasks and supporting more complex decisions.
But using AI and transforming with AI are two very different things.
EY’s 2025 Work Reimagined Survey found that 88% of employees are already using AI at work, yet only 5% are considered advanced users. The gap points to an opportunity business leaders are missing, determining how to put AI to work in ways that meaningfully change how the business operates.
That challenge is even more important as AI evolves from being just a tool into a technology that makes decisions on your behalf. Generative AI was once introduced to organizations as systems that could create, summarize, analyze, and assist. Agentic AI now takes the next step.
Amazon Web Services (AWS) defines agentic AI as autonomous AI systems that can act independently to achieve predetermined goals. Rather than waiting for a person to direct every interaction, AI agents can work toward defined objectives, make decisions based on context, take appropriate actions, and adapt as conditions change.
During the AWS Events session A Leader’s Guide to Advanced Team Structures in an Agentic World, AWS Executive in Residence Stephen Brozovich captured the organizational implications of that shift. As he explained, the conversation is no longer simply about if an organization should use AI. Leaders now must consider how they build teams that operate in this new environment while creating the structure, governance, and protections needed to embrace what comes next.
For contact centers, that shift has significant implications. Given the right context, knowledge, tools, and guardrails, AI agents can help move an interaction from intent to resolution, while bringing in a human when needed.
But the ability to act does not guarantee the ability to create value.
Consider a talented new employee entering the contact center. Capability alone is not enough to make that employee successful. They need to know the purpose behind their work, have access to accurate information, learn the processes they are expected to follow, receive feedback, and understand when to escalate an issue.
AI agents need the same kind of foundation.
For organizations, that makes agentic AI as much an operating model challenge as a technology opportunity. For contact center leaders, this creates an opportunity to rethink how customers, employees, and AI work together and how Amazon Connect can provide the foundation for a more intelligent, agentic operating model.
Unlocking that transformation starts with giving AI a clear purpose.
Agentic AI can determine next steps and act on its own, but why does that matter in the contact center?
The answer starts with what customers increasingly expect: resolution. Recent Zendesk research found that customer service agents view faster resolutions as 1.3 times more important to customers than reaching a live agent.
For contact center leaders, that raises the bar for AI. The opportunity is not simply to make interactions more conversational, but to help move customers toward resolution with less effort.
To do that effectively, AI agents operate through a continuous cycle of perceiving, reasoning, acting, and learning. An agent gathers information about the situation, interprets that context, determines and executes an action, and learns from the outcome to guide what happens next. Rather than relying solely on a predetermined sequence, the agent can respond to the conditions it encounters along the way.
The goal, however, is not autonomy for autonomy’s sake. It is applying autonomy where it can create meaningful value for customers and businesses. When designed around the right use cases, agentic AI can create opportunities for faster resolution, lower customer effort, and more efficient operations, while freeing human agents to focus on interactions where empathy and expertise matter most.
AWS Executive Tom Godden has described the “why” as essential to agentic AI. The more context an agent has around why an outcome matters and what it is expected to accomplish, the more effectively it can operate with greater autonomy.
AI agents need a clear purpose, but they also need access to trusted data, relevant knowledge, established processes, and clear parameters to interpret a situation and determine what comes next.
In that sense, agentic AI demands more from the foundation around it and uses the information and direction it receives to decide how to move work toward its given purpose.
Success, then, depends not only on where AI is deployed, but on how well it is equipped to perform. To invest with purpose and avoid unnecessary complexity, business leaders can approach AI agents much like members of a high-performing team.
Building a high-performing team takes more than bringing talented people together. Success depends on what surrounds them: clear goals, reliable information, established ways of working, ongoing development, and accountability. Building an effective AI team requires the same intentionality.
Every strong team understands the role it plays and what success looks like. AI agents need that same clarity.
This is why organizations should resist deploying AI simply for automating their system or to follow trends. Start with the business need. Is the goal to resolve a specific type of customer request? Reduce the effort required of human agents? Eliminate a recurring source of friction?
A well-defined role helps organizations determine what information, tools, permissions, and guardrails an AI agent actually needs. Instead of adding AI to an existing process and searching for value afterward, teams can design the agent around an outcome that matters from the start.
Employees rely on institutional knowledge to do their jobs, but much of that knowledge was created for people to read, not for AI to use.
Preparing information for AI requires organizations to look closely at the content behind their customer experiences. Knowledge should be accurate, current, and machine-readable, with metadata that helps establish context and relevance. Just as importantly, content needs clear ownership so there is accountability for keeping it useful over time.
For a contact center, that might include product information, policies, troubleshooting guidance, customer history, or other knowledge an agent needs to understand the situation and determine an appropriate response.
The quality of an AI agent’s decisions will always depend, in part, on the quality of the information available to it.
Much of how work gets done inside an organization lives in processes people have learned over time. Some are documented. Others depend on institutional knowledge, manual handoffs, or employees knowing which system to open next.
Agentic AI creates a reason to make those processes more explicit.
Organizations can translate repeatable procedures into workflows that AI agents can navigate and connect those workflows to the systems required to complete the work. In a contact center, that could mean moving beyond identifying why a customer is calling to knowing the steps required to update an account, check an order, initiate a return, schedule an appointment, or escalate an issue.
This creates procedural memory: a reusable understanding of how to carry out a task step by step. While semantic memory represents what an agent knows, procedural memory represents what an agent can do. Instead of determining the process from scratch each time, the agent can apply an established routine to move the work forward.
High-performing teams get better through coaching, feedback, and experience. AI operations should be approached with the same mindset.
Interaction data can reveal where knowledge is missing, workflows create friction, or an AI agent repeatedly struggles to complete a task. Those insights give teams a way to identify patterns and improve the system as customers and the business needs to change.
The important distinction is that improvement should be intentional. AI performance should not be treated as static, nor should organizations assume agents automatically become more effective simply through greater use. The operating model should create deliberate feedback loops for evaluating performance and making improvements over time.
Governance should not be something added after AI agents are deployed. It should be part of how the AI team is designed and managed from the very start.
That includes visibility into performance and behavior, controls that reflect organizational policies, and human involvement where it adds value. As AI takes on more work, governance provides the structure organizations need to expand its role responsibly.
The result is not an AI workforce operating separately from people. It is a model in which human expertise and AI capabilities are designed to complement one another. Once that foundation is in place, organizations can begin turning agentic AI from an idea into an operating capability across the contact center.
Once organizations have established the right foundation for agentic AI, the next challenge is putting those principles into practice at scale.
Amazon Connect Customer provides a cloud-based, AI-powered platform for doing just that. Built to support customer experiences across the journey, the platform gives contact centers a foundation for connecting customer context, AI, workflows, and human expertise.
Amazon Connect’s cloud-native design also gives contact centers greater flexibility in how they build and evolve their customer experiences. Organizations can configure capabilities around their needs and adapt as those needs change. A usage-based pricing model can also help organizations align spending more closely with demand, particularly as interaction volumes fluctuate between peak and off-peak periods.
But the opportunity goes beyond modernizing contact center infrastructure. With agentic AI, organizations can begin putting the operating model described above into action in three important ways.
Amazon Connect AI agents can engage with customers across voice and digital interactions, draw on relevant information and context, reason through the request, and take authorized actions to help move the interaction toward resolution. Powered by Amazon Bedrock, these capabilities create opportunities to automate more complex customer journeys that previously required multiple steps, systems, or human intervention.
For customers, the result can be a more direct path from “I need help” to “it’s handled.”
One of the most valuable applications of agentic AI is not replacing human agents but making them more effective.
When an interaction requires empathy, judgment, expertise, or additional authority, AI can support the employee taking over. Instead of asking agents to search across multiple systems and piece together context themselves, AI can help surface relevant information, summarize interactions, recommend next steps, and reduce some of the manual work surrounding the conversation.
This creates a more collaborative model in which AI handles the work it is well suited to perform, while people focus their attention where human capabilities create the greatest value.
Amazon Connect Customer provides access to enterprise AI capabilities alongside controls for security, privacy, and responsible AI use. Contact centers can establish what information AI agents can access, what actions they are permitted to take, and when an interaction should involve a person.
The objective is controlled autonomy by giving AI enough authority to create value while maintaining the oversight and boundaries the business requires.
Agentic AI creates new possibilities for the contact center, but the right technology requires the right implementation strategy to come together.
As an Amazon Connect Customer Ready Partner, USAN helps organizations design, implement, integrate, and optimize customer experiences on Amazon Connect Customer, turning the platform’s capabilities into solutions built around specific business and customer needs.
That approach can be seen in USAN’s work with OpenLoop Health. To modernize its patient experience, OpenLoop worked with USAN to implement capabilities across chat, email, voice, advanced IVR, and video. The organization also gained real-time visibility into operational metrics, enabling faster decisions and stronger SLA management.
Open Loop’s experience demonstrates the value of building the right foundation first. One that can evolve as customer expectations, business needs, and AI capabilities change. As organizations look to bring agentic AI into that environment, the same principle applies: successful AI transformation depends on more than deploying AI.
Whether you are evaluating your first AI use case or looking to scale AI across the contact center, USAN can help you build that foundation with Amazon Connect Customer and turn agentic AI potential into measurable customer and business outcomes.