Why AI Raises the Bar for CX Research

What the Next Era of Customer Intelligence Actually Demands

The question circulating in boardrooms, analyst briefings, and industry dossiers is: How much research will CX still need in the future? It is the wrong question. The right one is: What kind of intelligence will CX require — and are we structurally capable of delivering it?

The framing that CX budgets are shrinking because trackers show little movement misidentifies the symptom as the disease. What organisations are defunding is not CX research. It is measurement theatre: the ritual production of dashboards that confirm what everyone already suspects, at intervals that make the data irrelevant by the time it reaches a decision-maker. As Forrester’s Budget Planning Guide for CX Leaders noted, CX leaders face mounting pressure to demonstrate ROI from every research Euro — precisely because the old measurement model no longer earns its seat at the table.

That is not a research problem. That is a relevance and preparation problem. And confusing the two is expensive.

1. AI Raises the Bar. It Does Not Lower It.

CX has been shaped by AI-driven tooling for more than a decade. Early sentiment analysis and churn prediction models emerged in the early 2010s. Real-time personalisation engines followed. Conversational AI rewrote first-contact service architecture. Now, large language models are embedded in everything from customer-facing assistants to internal knowledge systems. Salesforce’s State of the AI Connected Customer, surveying 15,015 consumers across 18 countries, found that 63% believe advances in AI make trust even more important — a signal that each wave of AI deployment creates new research obligations, not fewer.

The assumption spreading through the industry — that AI will eventually answer the customer questions we used to pay agents and researchers to answer — is seductive and wrong at least for now. What AI does exceptionally well is process behavioural data at scale and speed no human team can match. What it cannot do is generate the right questions and solve issues in both B2B and B2C. That remains a human responsibility, and it requires more strategic discipline than ever as human interaction.

Gartner predicts that agentic AI will autonomously resolve 80% of common customer service issues by 2029. I am naturally sceptical about bold statements that Gartner does but let’s wait and see. This will not eliminate the need for CX research — it will radically transform its focus. The questions shift from “why did the interaction fail?” to “what trust architecture must AI-mediated service sustain?” That is a harder question, requiring deeper research capability.

McKinsey’s research on AI-powered customer interaction makes clear that the most accurate model still fails without human framing of what the AI is optimising for. Generative AI opens entirely new research territory: How do customers experience AI-mediated conversations differently from human ones? What happens to trust when empathy is synthetic? Legacy CX trackers were not designed to answer any of this. New methods, new frameworks, and more research are required — not less.

2. Journey Mapping Is a Museum Piece. Journey Management Is the Operating Model.

The customer journey map was the flagship artefact of CX strategy for two decades: colour-coded, cross-functional, pinned to conference room walls — and almost always static. Journey maps captured how things were at the moment of research. By the time they influenced decisions, the terrain had already shifted.

McKinsey’s landmark “From Touchpoints to Journeys” demonstrates that organisations managing end-to-end journeys dynamically — rather than optimising individual touchpoints in silos — achieve measurably stronger loyalty and revenue outcomes. Journey Management is where that logic lands: a continuous, cross-functional, data-fed operating model treating the customer journey as a live system, not a periodic diagnosis. Eglobalis’s analysis of Journey Management Architecture documents why the companies winning today have already made this structural shift.

This is already being operationalised at enterprise scale. SAP, Oracle, and Salesforce are converging around real-time journey orchestration. Adobe Experience Platform is building the infrastructure layer for this at scale. NTT’s Global Customer Experience Benchmarking Report, covering over 1,300 organisations across 34 markets, consistently finds that companies with integrated CX architectures significantly outperform those running disconnected measurement programmes.

The customer journey is not disappearing. It is becoming the central operating nervous system of the enterprise. That demands a research function capable of feeding it continuously — not one producing static artefacts quarterly.

3. The Metrics Layer Is Overdue for a Rebuild

Let us be direct: NPS, as a primary strategic metric doesn’t worked, has outlived its explanatory power for most organisations. Bain & Company — NPS’s own architect — has been evolving its application for years through NPS Prism and supplementary diagnostic frameworks, precisely because a single loyalty score cannot carry the strategic weight of a complex, multi-channel, AI-mediated environment. For a structural critique of why NPS fails in practice and many companies already realized it, see Why NPS Doesn’t Work Any More on Eglobalis.

In 2025 Global Customer Experience Index, tracking 275,000 customer perceptions of 469 brands across 12 industries and 13 countries, found that 73% of brands showed zero CX improvement year-over-year — while emotional quality remained the strongest and most underserved loyalty driver. A single NPS figure detects none of that.

KPMG’s Six Pillars of Customer Experience Excellence reinforces this: empathy, integrity, and personalisation — none of which a standard tracker measures directly — are the highest-order loyalty predictors across every market studied.

The future of CX measurement is a portfolio: behavioural signals (what customers do), predictive scoring (what they will do next), emotional texture (how they feel at decisive moments), and economic linkage (what each is worth in revenue). IDC projects that AI spending on personalised customer experiences will exceed $30 billion by 2027 in Asia-Pacific alone. Organisations anchored to NPS as a North Star are navigating by a map that predates the terrain. Luckily today we have hundreds of great data, and metrics to delivery a real 360-degree views of your customers

4. Friction Is Not the Problem You Think It Is

A widespread but dangerously oversimplified belief still dominates too many CX conversations: friction is bad, so remove it. That is not strategy. It is a reflex. The truth is more demanding. Friction can damage an experience, but it can also protect trust, reduce risk, create confidence, and force better decisions when designed properly. The real question is not whether friction should exist. The real question is which friction destroys value, and which friction protects it.

The foundational research is unequivocal. Dixon, Freeman, and Toman’s 2010 Harvard Business Review study “Stop Trying to Delight Your Customers”, based on over 75,000 customer interactions, established that reducing customer effort — not eliminating all friction — is the true driver of loyalty. 96% of customers who experienced high-effort interactions became more disloyal. The lesson is not “remove everything”; it is “remove the right things.”

Some friction is the product itself. The deliberate pause before a consequential financial transaction. The authentication step that prevents identity fraud. The design choice that slows a user just enough to ensure informed consent. Samsung’s AI strategy for customer experience, covering billions of touchpoints across hardware, software, and services globally, explicitly addresses the tension between frictionless convenience and user safety.

As the ECXO’s research on agentic AI and governance argues, the organisations leading in CX are not those eliminating all friction — they are those that govern which friction to eliminate. The CX research function must lead this distinction, classifying friction as costly, neutral, or purposeful.

5. What This Looks Like in Practice

Consider a large European financial institution that migrated a significant portion of its advisory process to AI-assisted interaction. Its CX tracker showed stable satisfaction scores throughout the transition. Leadership read this as validation. Qualitative research told a different story: customers rated interactions as “fine” — but were quietly consolidating assets with competitors they perceived as more transparent about when they were engaging with AI versus a human advisor.

The metric measured satisfaction. The research uncovered a trust erosion that would not have appeared in the score until the cost of reversal had become structural. This is precisely the intelligence gap that will define competitive outcomes over the next five years.

Medallia’s 2026 State of Customer Experience Report quantifies the gap: 66% of brands believe CX is improving — yet only 17% of consumers agree. That 49-point divergence is not a data problem. It is a research methodology problem: organisations relying exclusively on continuous measurement miss the contextual intelligence that explains why the gap exists.

Qualtrics XM Institute’s State of CX Management 2024 confirms that organisations with mature mixed-method programmes — combining continuous tracking with targeted qualitative and behavioural studies — identify loyalty risks significantly earlier and act on them more decisively than those relying on trackers alone.

6. Budget Transformation, Not Budget Reduction

Organisations cutting CX research or basic CX budgets are not becoming smarter. They are becoming more exposed. What is being rationalised is the research that was never decision-grade: tracker fatigue, survey overload, insight reports that generated no decisions.

BCG’s 2025 research on AI and customer experience argues we are entering a golden era for CX — but only for organizations that concentrate investment in decision-grade insight. Companies combining AI agent deployment with deep customer intelligence programs, such as Sandsiv, one of the fastest and most agile companies in our sector today, are already focusing on driving this smart AI evolution. With their ability to adapt faster than many larger players, they are beginning to achieve measurable competitive separation.

Salesforce’s 2026 State of Service, surveying 3,075 service professionals globally, found that AI agent adoption grew 1.7x in one year — and the #1 improved KPI was customer satisfaction, not operational efficiency. The organisations making this work are investing more in understanding customer context, not less. The investment is not shrinking. It is maturing.

7. Things Every CX Leader Must Do Now

Audit your measurement portfolio for decision-rate. Identify which metrics are generating operational decisions and which are generating reports. Redirect budget ruthlessly from the latter toward behavioural and predictive analytics.
Transition from Journey Mapping to Journey Management. Treat the customer journey as a live system requiring continuous intelligence, cross-functional governance, and real-time data integration. The map on the wall is not the territory.
Set your AI research questions before deploying AI tools. AI amplifies the quality of your questions, not the absence of them. Define your insight agenda first; then build the infrastructure to serve it. Without this, AI accelerates the production of irrelevant answers at scale.
Retire NPS as a standalone North Star. Build a portfolio of behavioural, emotional, and economic signals calibrated to your specific business model and customer architecture. One number is not a strategy — it is a shortcut.
Build a friction taxonomy before eliminating friction. Classify experience friction into three categories: costly (eliminate), neutral (monitor), and purposeful (protect). Optimising all friction downward destroys trust under the banner of convenience.
Invest in qualitative depth alongside quantitative breadth. The questions that will define competitive outcomes in the next three years cannot be answered by surveys alone. Data, behavioural observation, ethnographic research, and longitudinal studies are the intelligence layer AI cannot replicate.
Link every research commission to a named business decision. If you cannot identify the specific decision a piece of research is designed to inform, do not commission it. CX intelligence earns its budget by being indispensable to the P&L conversation — not by being voluminous.