Design a better customer experience with agentic AI

Whereas large language models (LLMs) attracted consumers well before enterprise users, SaaS platforms are driving the use of embedded AI agents to improve employee experience and productivity. AI agents are changing the future of work by helping HR leaders recruit personnel, marketers personalize ad campaigns, and IT service desk professionals respond to helpdesk tickets.

Building on this trend, the question is when—not if—embedded AI agents will become the new de facto standard for customer experience (CX). Instead of clunky user interfaces, complex search tools, and long data entry forms, businesses will simplify customer interactions with AI agents that know their preferences.

“Businesses seeking AI-driven value in customer experience will deploy specialized AI agents with domain expertise in product lines, inventory, pricing, delivery, and legal constraints,” says John Kim, CEO of Sendbird. “This shift is already transforming industries like retail, where AI enhances shopping through personalization and proactive service. In the future, consumers will have personal AI assistants—or multiple agents for finance, entertainment, healthcare, and travel.”

Start with the most boring, repetitive tasks

Some brands that pushed AI-enabled CX capabilities too early ran into issues impacting customers and their brands. Many organizations are taking a cautious approach to using AI agents to enhance customer experience, focusing their efforts on key prerequisites such as AI governance, data quality, and testing.

Early opportunities may be found in identifying boring use cases with a narrow scope, and with experiences that frustrate customers and occur at scale.

“The biggest impact that genAI and agentic AI can create is automating the most tedious, repetitive CX micro-workflows,” says Dave Singer, global VP at Verint. “A range of CX tasks, such as asking the right questions at the start of a customer interaction for context, searching for answers to customer questions, and conducting post-call wrap-up, can be automated with specialized AI-powered bots to drive stronger, faster business outcomes. As a result, agents have more capacity, CX is enhanced, and either cost savings, revenue generation, or both are increased.”

One opportunity is to consider the documentation and tools customers use to learn, install, and troubleshoot the product. Using an AI agent to answer questions can be a faster and easier experience than asking customers to find and read through many pages of documentation.

“Think about the ‘watering holes’ customers go to in their journey to use your product, like product help pages, user wikis, and online communities, and how genAI and LLMs can make those better and improve the customer’s experience,” says Jon Kennedy, CTO of Quickbase. “Make your solutions easier to use and more effective with a library of pre-built templates that can be accessed with a few prompts, all in the service of customers by industry, role, or other segments. Think about the ongoing education of those customers and how AI can help lead them to the next milestone to improve the customer journey, using the combined experiences built up in your customer community.”

Deon Nicholas, founder and president of Forethought, says to look beyond surfacing information faster and find easy ways to handle simple customer tasks. He says, “One of the easiest user experiences to develop with LLMs is a chatbot that delivers RAG-based search and can quickly surface information from FAQs in response to customer questions. However, embedding agentic AI into web and app experiences can deliver even better ROI because it takes action on behalf of customers, such as resetting their passwords or checking order status.”

Centralize access to customer data

Using AI in more interactive customer experiences requires using data that has been centralized and cleansed to train and validate AI agents. Companies use tools like customer data platforms and data fabrics to connect customer data and interactions.

“A robust AI-powered CX strategy is only as good as its underlying data and related governance, and any CX program should emphasize continuous testing and learning strategies to ensure data freshness and accuracy,” says Tara DeZao, senior director of product marketing and customer engagement at Pega. “This approach not only boosts agent performance but also mitigates risk and promotes brand equity as consumers interact with businesses across channels.”

When centralizing customer data, leaders must establish data controls around security, user access levels, and identity management. Many organizations utilize data security posture management (DSPM) platforms to reduce risks when they manage many data sources, multiple cloud database platforms, and disparate infrastructure.

“By rethinking how data is stored and accessed, moving from siloed third-party systems to user-centric data models, organizations can create more fluid, responsive web and mobile interactions that adapt to preferences in real-time,” says Osmar Olivo, VP of product management at Inrupt. “To maintain accuracy and performance, AI-driven experiences should be trained with diverse, real-world data while also incorporating user feedback mechanisms that allow individuals to correct, refine, and guide AI-generated insights by supplying their own preferences and metadata.”

Manish Rai, VP of product marketing at SnapLogic, predicts more than 80% of generative AI projects fail due to data connectivity, quality, and trust issues. “Success depends on tools that simplify agent development, make data AI-ready, and ensure reliability through observability, evaluation for accuracy, and policy enforcement.”

Rosaria Silipo, VP of data science evangelism at KNIME, notes many agentic applications have a human-in-the-loop step to check for correctness. “In other cases, special guardian AI agents focus on controlling the result; if the result is not satisfactory, they send it back and ask for an improved version.” For more data-related tasks, such as sentiment analysis, “genAI accuracy is compared to the accuracy of other classic machine learning models.”

Elevate customer service calls and chats to AI agents

Beyond information searches and fulfilling simple tasks are the customer service calls and chats that frustrate both consumers and the human agents serving them. In one survey, 23% of respondents claimed they would rather watch paint dry than go through repeated bad customer service experiences.

Instead of a rule-based chatbot with limited capabilities, customer service AI agents can sift through the data and respond to customers, while human agents tackle more challenging cases with the assistance of a customer success AI agent.

“There is a clear connection between customer satisfaction and the use of effective self-service,” says Vinod Muthukrishnan, VP and COO of Webex Customer Experience Solutions at Cisco. “The evolution towards truly agentic AI transforms self-service experiences by orchestrating end-to-end engagement between the brand and the customer. This advanced AI capability empowers customer experience teams to offer intelligent, seamless interactions, meeting customers where they are and on their schedules.”

Part of the challenge is data, and the other is that many customer experiences were developed as point solutions addressing only part of the customer journey. Technologists should apply design thinking approaches to re-engineer more holistic experiences when transforming to genAI-enabled engagements.

“Customer-facing applications such as websites, mobile apps, and B2C messaging typically have back-end integrations with customer-specific data sources that enable the application to answer questions and resolve problems,” says Chris Arnold, VP of customer experience strategy at ASAPP. “Leveraging an LLM to curate a personalized experience in a conversational format is far superior to the transactional experience offered by these applications alone.”

Fully test AI agents before deploying CX capabilities

Organizations looking to develop more advanced CX capabilities and autonomous AI agents will need a comprehensive testing plan to validate capabilities. Prompt filters, AI response moderation, content safeguards, and other guardrails help CX agents avoid conversations that are inappropriate or out of scope. However, brands must look beyond these basics and ensure that CX AI agents respond appropriately, accurately, and ethically.

“You can’t just throw an agent out there untested and unmonitored,” says Miles Ward, CTO of Sada. “Rigorous testing for accuracy and performance is non-negotiable. You need to know they’re delivering a frictionless, reliable experience, or you’re just creating a new set of problems.”

Ganesh Sankaralingam, data science and business analytics leader at LatentView, says AI experiences and LLM responses should be tested for accuracy and performance across five dimensions.

• Relevance: Measures how well the response is pertinent and related to the query.
• Groundedness: Assesses if the response aligns with the input data.
• Similarity: Quantifies how closely the AI-generated response matches the expected output.
• Coherence: Evaluates the flow of the response, ensuring it mimics human-like language.
• Fluency: Assesses the response’s language proficiency, ensuring it is grammatically correct and uses appropriate vocabulary.

Deon Nicholas of Forethought says, “Businesses should test AI experiences for accuracy and performance by running AI agents against historical customer questions and seeing how they do; measuring how often AI can handle customer interactions autonomously; and applying a separate evaluator model to check for conversation sentiment and accuracy.”

The future of AI agents in customer experience

How will AI agents impact CX in the near future? Mo Cherif, senior director of generative AI at Sitecore, recommends rethinking the experience entirely. “To create a genuinely agentic experience, don’t just enhance what already exists—build the journey specifically as a genAI-first interaction.”

There are contrary views on how AI agents will evolve. Some predict a more autonomous future where people empower and trust AI agents to make more complex decisions and take on a greater scope of actions. Others forecast a more human-centric approach, where AI agents augment human capabilities and partner with people to make smarter, faster, and safer decisions.

Michael Wallace, Americas solutions architecture leader for customer experience at Amazon Web Services, says agentic AI can resolve issues without human intervention. Consider a contact center that self-heals during a crisis, automatically redistributing resources, updating customer communications, and resolving issues before customers even experience them.

Wallace says, “Imagine an airline is facing a sudden traffic surge due to weather delays. With agentic AI, the contact center would make autonomous decisions about passenger rebooking and proactive notifications, allowing human agents to focus on complex customer needs rather than administrative tasks.”

Doug Gilbert, CIO and chief digital officer of Sutherland Global, says AI isn’t about automating customer experience; it should be about making experiences more human and intelligent. He says, “The true value of generative AI lies not in replacing human interactions but in enhancing them to be smarter, faster, and more natural. The secret is AI that learns from real-world interactions, constantly evolving to feel less robotic and more intuitive.”

Both autonomous and human-in-the-middle CX AI agents will likely materialize. In the meantime, businesses should fully research customer needs, improve data quality, and establish rigorous testing practices.