True Context Requires Data from Every Customer Touchpoint
The focus in too many analyst conversations and product roadmaps is on model capability – what features shipped this week, which frontier model powers what. But the companies that win won't be measured on features; they'll be measured on customer satisfaction, says Victor Belfor, Global VP, Business Development & Strategic Partnerships, 8x8.
Topics
As companies race to build AI-powered customer experiences, a deeper divide is emerging between those chasing features and those fixing infrastructure.
A recent Virtusa report found that 86% of enterprises are still treating customer data as operational residue rather than the foundation for real-time decision-making. The result is a widening gap between businesses that deploy AI at scale and those that merely experiment with it.
For many organisations, the new challenge is whether their systems can deliver accurate, connected information when customers actually need help. In that environment, integration is becoming the real differentiator.
“‘360-degree view’ has been beaten to death and doesn’t mean anything anymore. The question is whether the AI has the right information at the moment the customer reaches out, not five steps later,” said Victor Belfor, Global VP, Business Development & Strategic Partnerships, 8×8.
In this interview, Victor discusses why ‘context-aware AI’ is often misunderstood, why integration matters more than model capability, and why the future of CX depends on connecting platforms, people, and data more effectively.
Excerpts from the interview;
Everyone says AI is transforming contact centres. What’s the uncomfortable truth about why these systems still fail at scale?
AI doesn’t fix bad CX – it exposes and amplifies it. If your stack is fragmented, your data is stale, or your handoffs are broken, layering AI on top simply lets you fail faster, at scale, with a friendly voice.
Technology cycles like this follow a familiar pattern. First scepticism – companies tried pre-LLM AI, but didn’t see returns and walked away. Then exuberance, where a lot of vendors are still stuck, treating AI like something you pour over an existing platform and call it transformation.
The systems failing at scale now are mostly the ones that walked straight into that wall. Great CX still requires what it always did: voice, digital, and people combined into something that actually works. That’s what the 96% CSAT we see from customers comes from – not any single model, but the system around it.
‘Context-aware AI’ is becoming a catch-all. When you strip away the marketing, what does true context actually require?
The more an AI knows about a customer, the better its answers get. That sounds obvious, but operationalising it is genuinely hard.
True context requires data from every system that touches the customer, is refreshed in real time, and accessible at the speed of the conversation. Contact centre, CRM, and agentic AI are the obvious sources, but you also need payment systems, ticketing, vertical systems of record, and transcripts of calls and chats.
Whether it’s a native solution or the result of a partnership – like our deep integrations with Microsoft, Capacity, Synthflow, Smarsh, and others – the requirements are still the same. Pulling all of that together cleanly and keeping it fresh is the actual engineering challenge nobody talks about.
“360-degree view” has been beaten to death and doesn’t mean anything anymore. The question is whether the AI has the right information at the moment the customer reaches out, not five steps later.
If the data already exists, why does AI still struggle to make better decisions in real time?
The breakdown is rarely a lack of data. It’s that the data is fragmented, stale, and locked in systems that don’t talk to each other.
A customer’s last three interactions live in the contact centre, their account info in CRM, their last payment in billing, their support history in Zendesk, and their last sales call in Gong.
To make a smart real-time decision, AI needs all of it now – not in a nightly batch, not after a six-week integration project, but at the speed of the conversation. Most enterprise architectures weren’t designed for that; they were built for reporting.
The other piece people underestimate is data quality. AI is a remarkably efficient way to scale your data problems. If your records are stale, your call dispositions inconsistent, or your knowledge base contradicts itself, AI doesn’t fix that – it amplifies it.
What’s one widely held belief about AI in contact centres that you think is fundamentally wrong?
That better AI equals better CX. It doesn’t. Better integration of AI equals better CX.
The focus in too many analyst conversations and product roadmaps is on model capability – what features shipped this week, which frontier model powers what. But the companies that win won’t be measured on features; they’ll be measured on customer satisfaction.
Frontier models are commoditised. Any competitor can ship the same feature two weeks later. What doesn’t get copied in two weeks is everything around the model: latency under load, carrier-grade voice quality, clean handoffs between AI and human agents, identity resolution across channels, and the data model the AI is reasoning over.
Additionally, a critical layer is the ability to tailor and customise through partnerships, such as 8×8’s partnership with Synthflow. That’s where sustained competitive advantage actually lives.
Can a single platform realistically deliver context-aware experiences, or is this fundamentally a problem no one vendor can solve alone?
Honestly, both – and that’s the whole point.
Anyone who says a single vendor can build everything you need is selling you something. CRM, EHR, payment, ticketing, fraud detection, vertical systems of record – no company wins all of those.
But anyone who says you can stitch together 15 best-of-breed tools and call it a customer experience hasn’t actually tried to run a contact centre.
The answer is a strong platform at the core – voice, digital, agent desktop, AI orchestration, the underlying data model – and a serious partner strategy for everything else. This is where key partnerships – such as ours with Microsoft, Synthflow, Capacity, Smarsh, and more – make such a difference.
The vendors who lose are the ones trying to do everything themselves, and the ones with no real platform underneath at all. The winners are the ones with a strong centre of gravity and an open door.
ALSO READ: Breaking the ‘Walled Garden’ Habit in Mobile Performance
