Everyone is investing in agentic AI to improve customer experience. Chatbots, virtual assistants, predictive campaigns – the ambition is genuine. So why does the experience still feel fragmented?
The answer isn’t the AI. It’s what the AI has to work with.
Most enterprise teams operate across a patchwork of platforms: CRM, CDP, contact centre software, campaign tools, analytics dashboards, support ticketing systems. Each holds a piece of the customer picture. None of them were designed to share it. And AI that operates in that environment — no matter how capable — can only ever be as good as the data it has access to and the systems it can act on.
That’s the contradiction at the heart of modern customer experience. You promise personalisation, but your architecture guarantees every team sees a different version of the same customer.
The cost is measurable
Fragmented infrastructure doesn’t just create a bad experience — it creates a business problem. Lost revenue from journeys no single team owns end-to-end. Engineering overhead eating months of roadmap time just to maintain integrations. Agent burnout from toggling between systems to answer a question that should take seconds. Compliance risk from customer data scattered across vendors.
And customers feel it. 90% expect an instant response regardless of which channel they use. The average cost of a live agent interaction sits at $6, compared to $0.25–$0.70 for an AI-handled one.
Why adding channels makes it worse
The instinct is to add more — more channels, more tools, more touchpoints. But data from Infobip’s platform shows that 97.7% of traffic now comes from brands running multiple channels simultaneously. The brands accumulating complexity instead of capability are the ones adding channels without connecting them.
Customers move fluidly between WhatsApp, RCS, email, and voice depending on context and need. RCS grew 311% globally in 2025, with North American volumes growing 70x after Apple added support in 2024. WhatsApp grew 314% over five years and posted 15% growth in 2025 on an already massive base. 91% of Conversational AI interactions on the platform took place on WhatsApp.
Four channels is now the most common configuration, accounting for 25% of all platform traffic. A customer who messaged on WhatsApp last week shouldn’t have to re-explain themselves when they call your support line today. But in most systems, they do — because those channels don’t share context.
What agentic AI Really Means for Customer Experience
There’s a meaningful difference between a chatbot and an AI agent. A chatbot can tell a customer their order status. An agent can check the order, see it’s delayed, proactively offer a discount on their next purchase, and schedule redelivery – all in the same conversation.
Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024. The shift isn’t about better scripts. It’s about AI that understands intent, accesses the systems it needs, makes decisions, executes actions, and follows through – across channels, across platforms, across the full length of a customer journey.
The conversation doesn’t end when the customer gets an answer. It ends when their problem is resolved.
Three things that have to be true
For agentic AI to work at enterprise scale, three layers have to operate as one system.
First, data. Every customer touchpoint – messages, transactions, behavioural signals – needs to feed a single persistent profile. Agents that start every conversation knowing who they’re talking to, what they’ve done, and what they’re likely to need next operate differently from agents working from a clean slate.
Second, orchestration. Decisions about which channel to use, when to reach out, when to escalate, and when to hand off to a human need to happen at machine speed — not by static rule, but by reasoning from context.
Third, channels. All of them, natively integrated, sharing the same customer data, managed from the same interface. Not connected through fragile middleware, but built as part of the same system.
The results are already there
Floward, the leading online flowers and gifts platform across MENA and the UK, unified customer support across WhatsApp, Instagram, and Messenger. On Valentine’s Day – their highest-demand moment – they handled 54,000 conversations in a single day with a one-minute average response time, 95% SLA achievement, and a 14% cost reduction.
Vero, Brazil’s largest ISP, digitalised its entire customer journey through WhatsApp and SMS – from billing notifications and technical visit scheduling to debt collection and payments. The result: a 16x return on investment and a 194% increase in debt collection returns.
Nissan Saudi Arabia deployed Kaito, a 24/7 WhatsApp assistant handling enquiries in Arabic and English. Within six months: a 138% increase in leads generated and a 34x ROI in revenue.
These aren’t edge cases. They’re what happens when the infrastructure underneath the AI actually works.
Where to start
You don’t need to replace everything at once. Most teams already run campaigns, collect leads, and manage conversations across channels. The first step is connecting those interactions into a single customer profile. From there, orchestration and automation compound what’s already working.
If you already run chatbots, you have the foundation. The shift from chatbot to agent doesn’t require starting over – it requires connecting the data and giving the AI the authority to act on it.
The infrastructure to do this exists – AgentOS by Infobip brings it all together. The question is whether your current stack is set up to use it
