For years, AI has promised efficiency, personalisation, and growth. Yet for many organisations, reality remains limited. AI initiatives launch with momentum, stall in pilots, and rarely become part of everyday operations. The numbers are here: 98% of companies are currently experimenting with AI, yet only 5% of AI projects truly succeed (according to Gartner and BCG). For a detailed definition of Agentic AI, see our Agentic AI Playbook. For a detailed definition of agentic AI, see our Agentic AI Playbook.
In customer-facing organisations, this gap becomes especially visible in fragmented journeys, rising service costs, and inconsistent AI-powered customer experience, a challenge highlighted in Salesforce’s perspective on modern customer experience and trust.
The issue is not the AI models or platforms themselves. It is the data foundation they rely on.
Most organisations still operate with siloed, incomplete, or inaccessible data. This makes it nearly impossible for AI to act consistently or intelligently across the customer journey. Execution challenges do exist, but they only surface after the underlying data fragmentation has limited what AI can do in the first place.
Most AI systems are designed to respond rather than to act. They answer questions, generate insights, or automate isolated tasks, but they do not take responsibility for outcomes. As a result, operational waste persists; customer journeys fragment, and AI remains an add-on rather than a capability.
This is why the conversation around AI changes. AI moves from experimentation to accountability.
Why AI Customer Experience Initiatives Fail to Deliver ROI
Despite widespread experimentation, only a small share of AI initiatives reach production or deliver sustained impact. Research from Gartner and Boston Consulting Group consistently shows that while nearly all organisations invest in AI, very few achieve measurable ROI.
According to Gartner, agentic AI means that intelligent agents act autonomously to plan, decide, and execute, and is expected to appear in 33% of enterprise software applications by 2028.
The root causes are structural:
- AI lives outside core business workflows
- Insights stop at dashboards instead of triggering action
- Automation lacks context and brand intent
- Responsibility remains with humans
Without ownership inside operations, AI does not scale, and value stays theoretical. To change this, companies must move from treating AI as a novelty to making it an accountable, outcome‑driving operational capability that acts with context and relies on unified, high‑quality data to deliver real impact.
What Changes with Agentic AI Customer Experience
Agentic AI introduces a fundamentally different model.
Instead of supporting individual tasks, agentic systems own outcomes within clearly defined guardrails. They observe context, make decisions, and act across connected workflows.
In practice, this enables agentic AI customer experience that supports decisions, executes actions, and adapts journeys in real time; a direction Salesforce outlines within its vision for AI agents and autonomous systems.
Within the Salesforce ecosystem, this means AI agents that:
- Operate directly inside their platform and products
- Understand customer context in real time using unified data we are giving them access to
- Execute authorised actions across systems within defined guardrails established by the customer, not just recommend next steps
- Express brand identity consistently at every interaction
It’s a new team member who can execute and automate those time-consuming yet non-meaningful tasks that employees must perform, and that, according to Salesforce, account for 41% of their time every day. Thus, what Agentic AI does should be thought of almost as a job description, reducing friction where possible.
This is where customer experience transformation becomes operational instead of experimental.
Making Agentic AI Tangible
To make this tangible, let's take a concrete example based on a fictional luxury brand, Maison du Cacao. A fictional company with concrete examples rooted in our Merkle expertise and experience, as well as with public domain data.
The objective is simple: demonstrate what happens when AI is designed not as a chatbot, but as a branded operational layer embedded directly into daily work aligned with Salesforce’s approach to trusted, embedded AI.
How Agentic AI Pays Businesses Back
Each example below shows how operational efficiency improves while customer experience strengthens. A combination McKinsey identifies as essential for capturing value in its research on the state of AI in business. In each example, those agents serve your customers, acting as their 24/7 assistant to improve customer satisfaction.
- A Digital Claims Assistant reduces resolution times from weeks to under 48 hours by guiding customers through compliant, empathetic workflows while significantly reducing manual intervention.
- A Digital Concierge designs personalised stays in the brand’s tone of voice, addresses accessibility and dietary needs, and increases trust, satisfaction, and conversion.
- A Digital Sales Agent detects hesitation during checkout, offers timely guidance, and generates GDPR-compliant signals that improve retargeting and reduce both digital and physical waste.
But this can go even further; 24/7 assistants can solve many different issues and take back those non-meaningful tasks mentioned earlier. Did you know that according to the World Federation of Advertisers, up to 60% of digital media spend is wasted due to poor targeting? This is where the Data Intelligence Agent comes in handy and transforms journeys and interactions into emotional signals, revealing patterns that traditional dashboards do not surface enabling continuous optimisation.
Across every use case, the pattern remains consistent: less friction, clearer decisions, better outcomes.
How Tone of Voice Impacts Trust, Conversion, and Compliance
One of our biggest realisations? The tone of voice isn't cosmetic; it's a multiplier of trust, clarity, and conversion.
87% of luxury customers prefer brands that offer unique, tailored interactions during the purchase experience (McKinsey, PWC, fashionnetwork.com). Salesforce research on the State of the Connected Customer shows that customers are significantly more likely to trust and engage with brands that deliver personalised, consistent interactions, especially when AI plays an active role in the experience.
- When AI sounds generic, trust erodes.
- When AI sounds inconsistent, brand value weakens.
- When AI lacks empathy, friction increases, particularly in service and high-consideration journeys.
This is why AI agents are the perfect way to connect all customer touchpoints seamlessly, from boutique to e-commerce, customer care to after-sales. Well-trained agents prove that:
- Tone reduces friction
- Tone increases compliance in guided flows
- Tone influences buying behavior
- Tone is the experience
Companies spend millions building a brand… yet many still let their AI sound like a generic bot. Agentic AI becomes exponentially more powerful when it speaks your brand's truth. And when it does:
- Guided workflows achieve higher completion rates
- Customers move forward with greater confidence
- Experiences feel coherent instead of fragmented
Tone of voice becomes a conversion and compliance lever, not a design detail.
Why the Platform Foundation Matters
Agentic AI only delivers value when it operates directly inside the systems where work happens. No matter the technology stack, four ingredients are essential for autonomous execution at scale:
- Unified, trusted customer data
AI agents can only act intelligently when they access real-time, consistent data across channels, not siloed fragments. - Connected, orchestrated workflows
Agentic AI must move beyond recommendations and trigger actions across service, marketing, commerce, and operations. - Enterprise-grade governance and safeguards
Transparency, security, permissions, and guardrails determine what an agent can do, where it can act, and how it stays compliant. - Real-time decision and action layers
To reduce friction and influence journeys as they unfold, AI must operate closely with operational systems, not on disconnected dashboards
When these foundations come together, AI shifts from a supportive tool to an accountable capability. Agentic AI becomes part of daily business operations, not an isolated experiment.
From Experimentation to Execution
AI that pays back is not louder or more complex. It is intentional, integrated, and accountable.
At Merkle, we want organisations to be among the 5% of AI projects that succeed and ultimately to change that statistic altogether: from 5% to 10%, 15%, 30%, 60%… and beyond.
Agentic AI marks a fundamental shift: from intelligence that advises → to intelligence that acts.
This means moving from isolated AI tools to connected systems that drive operational efficiency, brand consistency, and measurable business outcomes. For organisations ready to move beyond experimentation, agentic AI becomes the practical foundation for sustainable customer experience transformation.
The Agentic AI Discovery Session
To help teams move forward with confidence, not guesswork; our focused working session aligns ambition with action.
You leave with:
- A prioritised opportunity map of agentic CX use cases tied to business impact
- Alignment between business priorities and AI capabilities
- A clear plan to shift from experiments to execution inside your workflows
- One validated agent use case, feasible in your real operating environment
This is not a theory workshop. It’s a fast, expert‑led working session designed to turn intention into operational next steps.
Ready to explore what agentic AI can do for your organisation?
If this sparked ideas about where smart agents could make a difference in your organisation, let’s explore them together. Send a quick email to Kevin Kohnke, Senior Business Transformation Manager, to schedule your Discovery Session.
