Artificial intelligence (AI) has progressed from experimental pilots to an integral component of enterprise customer experience (CX) strategies. While many organizations now leverage AI agents for customer service, a significant gap persists between AI’s technological capabilities and its current operational deployment. This disparity often leads to consumer dissatisfaction and a missed opportunity for businesses to fully realize AI’s strategic value. Agentic CX in 2026: What consumers expect and most enterprises miss a recent global research study conducted by Ada and NewtonX, surveying 2,000 consumers and 500 CX decision-makers, reveals critical insights into consumer expectations, enterprise measurement shortcomings, and the organizational complexities hindering AI’s potential .
The research indicates that consumers are not inherently skeptical of AI; rather, they are critical of AI that fails to resolve their issues. Currently, only 24% of consumers report achieving full resolution from an AI agent without human intervention. Concurrently, 55% of businesses lack clear visibility into their AI agents’ performance, making sustained improvement structurally challenging. The path to advanced agentic CX requires a fundamental shift in how enterprises approach AI implementation, measurement, and governance.
The Resolution Imperative: Consumer Expectations for Agentic CX
Consumers demand efficient and effective AI, prioritizing concrete problem resolution over perceived “human-like” empathy. The data indicates a clear mandate for AI that works seamlessly.
The Current State of AI Resolution
Despite widespread AI adoption, many consumers encounter unresolved issues when interacting with AI agents. Only 24% of consumers report their issues being fully resolved by AI . The remaining interactions frequently result in negative outcomes: 33% are escalated to a human, 22% achieve only partial resolution, and 11% lead to complete abandonment . These failures are rooted in AI’s operational limitations, not a preference for human interaction. Key failure points include:
- Comprehension failures: 74% of consumers cite instances where AI agents looped or failed to understand their questions.
- Capability gaps: 56% encountered situations where AI could not handle the complexity of their request.
- Repetitive responses: 50% experienced AI agents repeating unhelpful information .
These issues often lead to greater frustration than if the consumer had reached a human agent immediately. As one Service Operations & Technology Department Head in Travel & Hospitality noted, “AI gets them 70-80% there, then fails. That makes the experience more frustrating than if they’d reached a human immediately” .
Prioritizing Capability Over Empathy
When asked to allocate points across various customer service features, consumers overwhelmingly prioritized core capabilities over empathetic communication. Top priorities included accuracy (18 points), problem-solving ability (18 points), and the option for human escalation (16 points), followed by data privacy/security and speed of resolution. Empathy ranked last (6 points) . This underscores that consumers seek effective solutions, not AI that mimics human sentiment. For instance, in a billing error scenario, consumers chose a straightforward, solution-focused AI that resolved their issue in 10 seconds over a conversational, empathetic AI that took two minutes . The core insight is straightforward: “Consumers aren’t anti-AI. They’re anti-bad customer service” .
Building Trust Through Transparency and Control
Consumer trust in AI is directly correlated with the complexity and potential consequences of the interaction. While AI is preferred for simple tasks like basic FAQs, preference shifts towards human agents for moderate tasks (e.g., account updates, status checks) and becomes pronounced for complex or urgent issues (e.g., financial advice, lost credit cards) . This highlights that complexity itself is not a barrier; rather, it’s a visibility problem concerning AI’s reliability in high-stakes scenarios.
What to do:
- Ensure immediate and transparent AI disclosure: 74% of consumers expect to know they are interacting with AI at the outset. This empowers them to decide whether to continue with the AI agent or request a human .
- Implement frictionless human escalation: 57% of consumers indicated they would stop using a company’s AI service if they could not transfer to a human when needed . Provide clear, accessible pathways for escalation without unnecessary friction or questioning.
- Focus on secure, API-driven resolutions: Emulate successful models like Tilt Financial Services, which achieved 84% AI-resolved chat volume and an 8-point CSAT lift by enabling AI to handle moderate to complex tasks through authenticated customer data access via secure APIs .
What to avoid:
- Forcing AI-first resolution policies: 41% of businesses mandate that AI agents attempt resolution first . This can lead to frustration and abandonment if the AI fails.
- Withholding AI disclosure: 10% of businesses do not disclose AI interaction, and 13% only disclose at the point of human handoff, removing consumer agency .
Summary: Consumer trust in agentic CX is built on the AI’s demonstrated capability to resolve issues accurately and efficiently, supported by transparent disclosure and accessible human escalation options.
The Measurement Imperative: Aligning AI Performance to Business Outcomes
While enterprises are investing heavily in AI, many are measuring outcomes based on traditional efficiency metrics, rather than the resolution-focused metrics consumers value. This misalignment impedes the ability to prove AI’s strategic impact.
The Misaligned Focus of AI Measurement
Most CX leaders prioritize operational efficiency when evaluating AI deployments. The top benefits cited by businesses include reduced wait/handle times (51%), lower cost per interaction (47%), 24/7 availability (43%), and ticket deflection (41%) . In contrast, critical customer-centric metrics such as resolution rate (22%) and CSAT improvement (20%) rank significantly lower . This indicates a fundamental gap between what businesses measure and what customers value.
Containment and deflection metrics, while useful for managing contact center volume, fundamentally measure avoidance of human interaction, not the successful fulfillment of customer needs. True “automated resolution” must signify end-to-end completion of an issue, accurately and in alignment with policy, without human intervention . Without this precise definition, “success” becomes ambiguous, corrupting business intelligence and hindering meaningful improvement.
Establishing a Robust Attribution Infrastructure
Over half of businesses (55%) lack the robust infrastructure required to accurately monitor AI agent performance. Many measure AI and human agent interactions together (blended metrics), or only by channel, making it impossible to isolate AI’s specific impact . Only 28% of businesses achieve the “gold standard” of tracking AI-only, hybrid, and human-only interactions separately . This lack of granular visibility prevents organizations from:
- Pinpointing AI performance gaps: Identifying where AI is failing or excelling.
- Establishing credible ROI baselines: Proving the direct financial return on AI investments.
- Systematically improving performance: Iterating and optimizing AI agents effectively.
The commercial case for AI scaling extends beyond initial efficiency gains. It requires demonstrating measurable impact on downstream business outcomes such as customer lifetime value (CLTV), retention, and incremental revenue. Currently, 42% of businesses cannot link AI interactions to downstream outcomes, and 32% cannot isolate AI’s impact from other CX improvements . The mandate has evolved from simply deploying AI quickly to providing tangible proof of its value.
Achieving Measurement Maturity
Advancing AI in CX necessitates a staged approach to measurement maturity:
- Level 1: No segmentation. All interactions are measured collectively, often through CRM defaults without agent-type tagging.
- Level 2: Partial segmentation. AI-only interactions are separated from human-assisted interactions, typically using unique ticket flags or manual sampling.
- Level 3: Full segmentation. Pure AI, hybrid, and human agent interactions are tracked separately. This requires automated CRM tagging and AI platform logging.
- Level 4: Attribution. AI’s impact is linked directly to downstream business outcomes such as CLTV and retention, often through A/B cohort testing over extended periods (e.g., 12+ months) .
What to do:
- Implement full segmentation immediately: Prioritize tracking AI-only, hybrid, and human interactions distinctly to gain granular performance visibility.
- Define “automated resolution” rigorously: Ensure internal definitions align with true end-to-end problem completion, not just contact avoidance.
- Develop attribution models: Link AI performance data to loyalty, retention, and revenue metrics.
- Leverage efficiency for strategic reinvestment: As demonstrated by Simba Sleep, which redirected three agents from routine inquiries to high-value interactions (abandoned carts, sales callbacks), generating approximately £600,000 per month in additional revenue . Similarly, IPSY strategically reinvests efficiency savings into elevating member experience, transforming customer care into a growth driver .
Summary: To move beyond efficiency metrics and establish AI as a strategic growth driver, enterprises must build a robust, multi-layered measurement infrastructure that provides clear visibility into AI performance and its direct impact on commercial outcomes.
The Governance and Talent Imperative: Scaling Agentic CX
Scaling AI beyond basic use cases requires more than just technology; it demands a robust governance framework, seamless system integration, and a skilled workforce capable of managing and optimizing complex AI deployments.
The Ambition-Capability Gap
While 92% of businesses intend to increase their AI investment in CX over the next year, only 9% are primarily AI-led today, with 21% expecting to be within 12 months (a 2.4x increase) . This ambition contrasts with the reality that most current AI deployments handle low-complexity tasks such as basic FAQs or simple how-to guidance (66-67% of businesses) .
To progress, AI agent scope must expand through maturity stages:
- Stage 2 (Developing): Read-only access to customer data (e.g., status checks, order tracking) and basic account updates (e.g., password resets) (36-54% of businesses).
- Stage 3 (Optimizing): Edit access to customer systems (e.g., credit card renewal, subscription changes, processing exchanges) and handling service issues/complaints (52% of businesses).
- Stage 4 (Innovating): Complex, multi-agent orchestration, human-in-the-loop scenarios, eligibility checks, and even providing consultation (e.g., insurance policy, financial advice) (14-24% of businesses) .
This progression highlights the need for AI to perform actions and automate back-end systems, moving beyond simple information retrieval.
Overcoming Key Barriers to Maturity
Several critical barriers prevent businesses from advancing their AI maturity:
- AI accuracy not sufficient (51% citing).
- System integration complexity (47% citing).
- Regulatory/compliance uncertainty (40% citing).
- Fears of reputational risk (34% citing).
- Budget/resource constraints (33% citing).
- Negative customer perception (32% citing) .
What advanced businesses do :
- Set explicit accuracy thresholds per use case before deployment.
- Conduct A/B testing to establish performance baselines before scaling.
- Prioritize API-ready use cases and ensure a middleware strategy for complex integrations.
- Align internal governance frameworks before deployment, including Role-Based Access Control (RBAC), and test use cases in a sandbox environment.
- Implement disclosure-first designs and ensure easy escalation to human agents when needed, measuring customer acceptance at each complexity tier.
Building the Organizational Foundation
Effective AI deployment hinges on robust governance and a skilled workforce. The study reveals that 80% of businesses lack a fully adopted, aligned governance framework for AI in CX . Governance, defined as the rules, controls, and data practices dictating AI behavior and its boundaries, is crucial for safely deploying AI in high-stakes environments. Among businesses with formal frameworks, 44% co-developed them with their AI vendors, highlighting the value of expert collaboration in anticipating potential gaps.
What to do:
- Establish a comprehensive AI governance framework: Define clear rules, controls, and data management protocols for AI agents. This framework should be developed cross-functionally and, ideally, in collaboration with AI vendors to preemptively address technical safeguards and compliance requirements.
- Invest in specialized AI talent: Address critical skills gaps in conversation design and dialogue flow (28%), system integration and API management (38%), and AI ROI modeling/business case development (38%) . Many CX teams (36%) are not adequately resourced or skilled to manage and audit AI effectively, rising to 62% in human-led organizations transitioning to hybrid models.
- Define a dedicated operating model and roles: Implement roles such as full-time ACX Managers for QA, coaching, and data analysis; dedicated Integration Engineers; and Knowledge Managers. Senior leadership and C-level buy-in are essential to integrate ACX strategy into the core business vision.
What to avoid:
- Delaying governance framework development: Without clear guardrails, businesses risk deploying AI in scenarios where errors can cause significant harm or reputational damage.
- Underestimating skill development: Assuming existing CX teams can manage AI effectively without specialized training or new hires.
Immediate priorities (first 90 days):
- Initiate AI governance working group: Assemble cross-functional stakeholders (legal, compliance, CX, IT, security) to draft a preliminary AI governance framework, focusing on data privacy, ethical use, and escalation protocols.
- Conduct a skills gap analysis: Evaluate current CX and IT teams against the requirements for managing advanced AI agents (e.g., conversation design, API integration, performance analytics). Begin planning targeted training or recruitment.
- Review system integration readiness: Assess current CRM, billing, and other core systems for API readiness and data accessibility to support expanded AI agent capabilities.
Summary: Achieving advanced agentic CX requires a strategic organizational commitment to robust governance, proactive system integration planning, and the cultivation of specialized AI talent, rather than solely focusing on technological deployment.
Summary
The future of agentic CX is not defined by the sheer volume of AI investments, but by the strategic quality of those investments. Consumers consistently seek AI that delivers accurate, efficient resolutions, coupled with transparency and accessible human support. Yet, many enterprises remain hampered by misaligned measurement strategies, underdeveloped governance, and critical skills gaps, preventing them from realizing AI’s full potential.
To transform customer experience from a cost center into a growth driver, CX leaders must pivot from merely deploying AI to building a robust foundation that ensures AI effectiveness and measurable business impact. This entails implementing sophisticated measurement infrastructures to track AI performance against commercial outcomes, establishing comprehensive governance frameworks to manage risk and ensure ethical deployment, and cultivating the specialized talent required to operate and optimize advanced AI agents. Organizations that prioritize these foundational elements today will build lasting competitive advantages in an increasingly AI-driven market.
Source: Ada & NewtonX. (2026). Agentic CX in 2026: What consumers expect and most enterprises miss. Ada.
