The customer-experience world is having its great divide moment. On one side, the legacy giants are still trying to patch AI into platforms built long before tools like ChatGPT existed. On the other hand, a new generation of AI-first CX companies are building from a clean slate, treating data, automation, and integration as the foundation, not an upgrade.
Few people know that tension better than Oscar Giraldo. The former Playvox CEO watched what happens when big players move slowly after NiCE acquired his company. “It’s been over a year since the acquisition, and they’re still not fully integrated.”
That frustration turned into inspiration for his next venture, Oversai: a full-stack AI platform built entirely for the new agentic era.
Across the market, the cracks are showing. NICE’s billion-dollar Cognigy deal shows that legacy vendors are scrambling to buy innovation they couldn’t build themselves. Meanwhile, most CX leaders say integration and data quality are their biggest AI headaches.
It’s starting to seem like the real winners won’t be those who talk loudest about AI; they’ll be the ones who built it right, from the ground up.
AI-First CX: From Retrofitted AI to Real Integration
If the rise of AI-first CX feels fast, it’s because the old guard is finally seeing just how far behind they’ve fallen. Giraldo doesn’t mince words when describing the difference between bolting AI onto old systems and building it in from day one.
“Companies announce partnerships with brands like Snowflake because they know they need data from everywhere – but they don’t own that layer.” Giraldo says. “Realistically, none of the legacy vendors started their companies after ChatGPT – they’re struggling to build a native data layer.”
This “retrofit trap” is what’s holding so many large CX platforms back. Their architectures were never designed for the data-hungry reality of modern CX automation. Instead, they stack new AI features on top of outdated databases and fragmented workflows, a structure that Giraldo compares to “installing smart bulbs in a house with no electricity.”
The numbers back him up. Gartner predicts that 40 percent of agentic AI projects will fail by 2027 because the underlying data systems can’t support real integration. When that happens, vendors lose trust. That’s where Oversai draws its line.
Giraldo’s new platform is built on a universal ontology: a shared data model that lets AI agents pull and act on information from CRMs, e-commerce sites, help desks, or claims systems in real time.
When asked why the industry feels stuck, Giraldo points to the acquisition frenzy. “If legacy vendors really want to compete, they need to acquire, that’s why NICE paid nearly a billion,” he says. “But you can’t buy modern architecture. You have to design it.”
Analysts agree. BCG estimates that organizations embedding agentic AI directly into core systems can accelerate business processes by 30 to 50 percent, while those relying on retrofits will see diminishing returns.
AI-First CX and The Agentic Enterprise
The shift from static automation to dynamic action is happening fast. For years, the industry has promised AI for customer experience, but most systems could only talk, not act. Now, a new generation of AI-first CX platforms is delivering on that promise, creating what Oscar Giraldo calls the agentic enterprise.
In Giraldo’s view, the next leap in CX automation isn’t about generating more conversation, but about enabling execution. Oversai, his new company, is built around the idea that an AI agent shouldn’t just answer a support query; it should resolve it from start to finish. “AI agents need access to everything, from inventory to claims systems,” he explains, “that’s the difference between AI that talks and AI that acts.”
That requires something legacy architectures were never designed for: true CX integration. Oversai connects systems through a shared ontology, which allows AI to retrieve, interpret, and act across multiple business tools. In practice, that means an OversAI agent could start a WhatsApp conversation, check stock in Shopify, generate an order in HubSpot, trigger an invoice, and post an update in Slack, all without a human in the loop.
Giraldo compares OversAI’s design to Tesla’s vertical integration: software, hardware, and batteries all built together for seamless performance. “We’re integrating from the data stack all the way to the agents on the front end,” he says, underscoring the advantage of full-stack AI platforms that own every layer.
The Need for a Full-Stack, Data-First Architecture
In today’s AI-first CX market, the real competitive edge isn’t a clever algorithm; it’s what lies beneath it. While every vendor has access to similar models, few control the quality, flow, and structure of their data. That, says Oscar Giraldo, is where the true moat exists.
“Oversai was born post-ChatGPT,” he explains. “We built the ontology and data stack from day one. Everyone has the same models; the moat is the data layer.”
It’s a sharp contrast to how most legacy platforms operate. Traditional systems rely on a patchwork of integrations, APIs, and middleware that pass data around but rarely unify it.
Oversai flips that model by building from the data layer up, allowing its AI agents to work seamlessly across applications. The result is a full-stack AI platform that controls every component without the friction of third-party dependencies and adapts to specific needs.
The benefits of that philosophy extend beyond just better automation. Oversai also includes workforce management and quality monitoring within the same architecture, ensuring companies can oversee not just what AI agents do, but how well they do it. “Now that AI agents are doing the work, you need oversight on how they perform; that’s why we built quality management in from the start,” Giraldo says.
That’s crucial at a time when companies need to balance the potential of AI with data governance, security, and oversight.