Sierra’s new Liberty Global partnership will bring AI agents to businesses serving roughly 80 million fixed and mobile connections across Europe. But the scale isn’t the most interesting part. Sierra is increasingly building agents that don’t just resolve customer-service interactions. They remember, pursue objectives over time, and generate measurable business outcomes. That could fundamentally change the economics of customer experience.
For decades, the contact center has operated under an uncomfortable assumption: customer conversations are primarily a cost to be managed.
The technology changed dramatically, but the economics remained remarkably consistent. Companies invested in better routing, workforce optimization, self-service, cloud infrastructure and automation, yet many of the industry's defining metrics still revolved around the same objective, that is to handle more customer interactions at a lower cost.
AI initially appeared to accelerate that model. If an AI agent could resolve an interaction that previously required a human employee, the economic proposition was straightforward: automate more conversations and reduce the cost of serving customers.
Sierra appears to be pursuing something more ambitious.
Sierra's newly announced three-year strategic partnership with Liberty Global will provide a framework for deploying Sierra's AI agents across Liberty Global operating companies representing approximately 80 million fixed and mobile connections in Europe. The agents will engage customers across chat, voice and text, with deployments already underway.
At that scale, the announcement is significant on its own. But viewed alongside Sierra's recent product development and customer deployments, a more consequential story begins to emerge.
Sierra isn't simply trying to automate the contact center. They are trying to change what the contact center is economically capable of doing.
From Resolving Conversations to Producing Outcomes
Most customer-service technology has historically been optimized around an interaction.
A customer calls. A conversation begins. The issue gets resolved, or it doesn't, and the interaction ends.
Even sophisticated AI agents largely inherited that architecture. They became significantly better at understanding questions, accessing enterprise systems and completing tasks, but the fundamental unit of work remained the conversation.
Rather than requiring an AI agent to accomplish everything within a single interaction, Horizon agents can pursue an objective across days, weeks or even months. They can respond to signals, determine the next action, engage customers across channels, retain context between interactions and continue working toward a defined business outcome. Sierra gives examples ranging from originating a mortgage and reducing churn to completing healthcare referrals and converting customers before a trial expires.
That sounds like a subtle architectural change.
Consider the difference between an AI agent answering a mortgage question and an AI agent being given the objective of helping a qualified borrower complete a mortgage application.
The first performs customer service.
The second performs work.
It may need to contact the customer again, recognize what documentation is missing, adapt when the customer stalls, move between SMS and voice, coordinate with enterprise systems and eventually determine when a human employee should enter the process.
At that point, the AI agent isn't simply replacing a conversation that would otherwise have been handled by a contact-center employee.
It's participating in the economics of the business.
Figure is using Sierra's Horizon platform to re-engage consumers who have stalled during home-equity applications. The agent works across voice and SMS over multiple days, helping borrowers through friction points such as credit-check permissions, identity verification and bank-account linking before transferring appropriate customers to a human loan officer.
The early results are notable.
According to Figure, stalled applicants interacting with the AI agent progressed through individual friction stages at rates 30% to 52% higher than those who did not. Borrowers engaging with the agent funded 67% more loan volume, while combining the AI agent with human loan officers produced a reported 143% lift in funded-loan conversion compared with loan officers operating alone.
Those numbers deserve continued scrutiny as deployments mature, and they come from the companies involved rather than an independent study. But the use case itself illustrates something important.
This isn't primarily a contact-deflection story.
It's a revenue story.
And Bret Taylor, Sierra's CEO and Co-Founder, is increasingly explicit about that distinction.
“Enterprise AI is increasingly moving beyond reactive, cost saving use cases to complex, revenue-generating opportunities.”
Bret Taylor
CEO and Co-Founder @ Sierra
The transition from saving money to producing economic outcomes could prove far more disruptive to the CX technology market than simply replacing portions of human customer service.
What Happens When an AI Agent Has Infinite Patience?
There is another characteristic of this model that is easy to underestimate.
Humans operate under enormous constraints.
A loan officer has hundreds of prospects. A customer-success manager manages dozens or hundreds of accounts. A retention team cannot personally follow every weak signal that suggests someone might leave.
AI doesn't face the same scarcity.
An AI agent can theoretically monitor millions of customer relationships simultaneously and decide that one customer needs attention today, another next Thursday and another shouldn't be contacted at all.
More importantly, it can wait.
A customer who abandons an application doesn't necessarily need another generic marketing email five minutes later. Perhaps they need help three days later when a specific document remains missing. A subscriber showing early signs of churn might require a completely different intervention from someone who has already tried to cancel.
Sierra's Horizon architecture is explicitly designed around this idea. Its agents can identify signals, retain context, decide what action comes next and continue pursuing an objective across multiple engagements rather than treating each conversation as a clean slate.
That creates an intriguing possibility:
AI may make economically viable millions of small customer interventions that companies could never afford to perform manually.
And that begins to expand the definition of customer experience itself.
The Contact Center Starts Looking Less Like a Department
For decades, enterprises built organizational boundaries around customer interactions.
Sales had its systems. Marketing had its systems. Customer service had the contact center. Customer success managed relationships after the sale. Operations executed workflows behind the scenes.
Customers never cared about those boundaries.
Agentic AI may begin erasing them.
Imagine an AI agent that can answer a billing question today, recognize emerging churn risk next month, proactively propose the right retention offer, help the customer upgrade six months later and remember the context surrounding every interaction along the way.
Is that customer service?
Sales?
Customer success?
Marketing automation?
Increasingly, the answer may simply be the customer relationship.
Sierra already describes the AI agent as potentially becoming a company's "digital front door," ultimately surpassing websites and mobile apps as a primary interface between businesses and customers. Horizon extends that concept by allowing the agent to continue working after the customer closes the chat window or hangs up the phone.
If that architecture succeeds, the traditional contact center begins looking less like a destination customers enter and more like an intelligence and orchestration layer operating continuously around them.
Customer Memory Becomes the Moat
But autonomous agents have a problem.
Intelligence alone isn't enough.
Two competing companies can increasingly access many of the same frontier AI models. If both organizations can rent comparable intelligence, the model itself becomes a difficult place to build durable differentiation.
What can't be rented as easily is the accumulated knowledge of a company's customers.
Sierra calls this its Context Engine. It combines customer history, enterprise information and the results of previous interactions so agents can determine what matters and what action is most likely to produce the desired outcome. As more interactions occur, the system can learn which interventions work for particular customers and situations.
This is where Sierra's strategy intersects with a broader shift that we have been watching across the industry.
Glia recently described its approach to banking CX around the idea of a “single memory” of the customer. Sierra talks about persistent memory and proprietary customer context. Different architectures and markets, but the strategic direction is remarkably similar.
The underlying AI model may become increasingly interchangeable.
Customer memory won't.
A competitor can license the same foundation model tomorrow. It cannot instantly reproduce years of knowledge about which customers churn, which offers work, how individual relationships have developed and what thousands or millions of previous customer interactions have taught the organization.
That accumulated context becomes proprietary intelligence.
And every successful interaction potentially makes the next one better.
Sierra Is Challenging the SaaS Business Model, Too
This is another reason Sierra deserves attention.
It isn't only changing what AI agents do. It's challenging how enterprise software gets paid.
Traditional contact center software largely monetizes access: seats, licenses, usage, consumption or some combination of them.
Sierra increasingly emphasizes outcome-based pricing.
Its argument is straightforward: if AI can actually perform work, customers shouldn't necessarily pay for how many tokens the system consumed or how long the software ran. They should pay for the outcome it delivered. Sierra says Horizon extends this philosophy to longer-running workflows: the customer pays for results rather than token consumption.
That creates an entirely different vendor relationship.
A CX software company traditionally sells technology that helps an enterprise produce an outcome.
An agentic AI company increasingly says: Tell us the outcome. The software will do the work.
That distinction could become uncomfortable for incumbent enterprise communication software companies because it shifts the conversation from software functionality toward economic value.
If an AI agent saves a $1,500-a-year customer from churning, what is that worth?
If it converts an abandoned mortgage application into a funded loan, what is that worth?
If it increases lifetime value, completes a healthcare referral or converts a trial into a paying subscriber, why should the vendor's economics necessarily be tied to seats or tokens at all?
Sierra isn't merely changing who performs the work.
It's challenging how the work itself gets valued.
Liberty Global Gives the Thesis Scale
This brings me back to Liberty Global.
The headline number, approximately 80 million fixed and mobile connections, is impressive. But the more important part may be what happens when this architecture is exposed to that many customer relationships.
Telecommunications is an ideal testing ground. Customers contact providers about billing, service problems, device upgrades, cancellations, plan changes, installations and countless other issues. Those interactions generate enormous amounts of information about satisfaction, intent and commercial opportunity.
Sierra's agents will initially handle routine and increasingly complex interactions across Liberty Global's businesses, while human care teams focus on situations requiring judgment and deeper expertise. But Taylor's description of the partnership is telling: he says "Sierra aims to improve customer experience, reduce support burden and grow revenue."
That third objective is where this story gets interesting.
If AI agents can continuously learn from customer interactions, retain context across channels and proactively pursue outcomes, telecom operators may eventually stop evaluating the technology primarily by how many calls it deflects.
They may start asking how much economic value each customer relationship produces.
The Next Contact Center May Not Look Like a Contact Center
None of this means human customer service disappears.
In fact, Sierra and Liberty Global explicitly position AI as allowing human teams to spend more time on interactions requiring judgment and expertise. Figure's early results are similarly interesting precisely because the strongest reported performance came from AI agents working alongside human loan officers, not simply replacing them.
The more interesting possibility is that the organizational boundary around the contact center begins to disappear.
AI agents don't particularly care whether an interaction belongs to service, sales, retention, onboarding or customer success. They care about the objective, the context required to pursue it and the actions they're authorized to take.
That could eventually force enterprises and CX vendors to rethink a category that has existed for decades.
The contact center was built around conversations.
Agentic AI is increasingly being built around relationships and outcomes.
Sierra's Liberty Global agreement provides enormous scale for testing that idea. Figure provides early evidence that it can translate into revenue. Horizon provides the architecture for agents that operate beyond individual conversations, while Sierra's Context Engine attempts to turn the resulting customer knowledge into a compounding advantage.
Taken together, Sierra's strategy raises a question that is much bigger than whether AI can automate another percentage of customer-service volume.
If AI agents can remember customers, operate across channels, pursue objectives for weeks or months and increasingly generate revenue, will we even describe the “contact center” as a distinct function five years from now?
Or will what we currently call the contact center evolve into an autonomous customer operating layer running continuously between the enterprise and its customers?
Sierra appears to be building toward the latter.
Stay updated with cx news
Subscribe to our newsletter for the latest insights and updates in the CX industry.
