For decades, the contact centre has been improving at answering the same fundamental question: How efficiently can we handle customer demand?
We built workforce models around it. Technology stacks around it. Outsourcing contracts around it. Dashboards around it.
Average handle time. Service level. First-contact resolution. Occupancy. Quality. CSAT. Cost per contact. The industry became highly sophisticated in managing the arrival, routing, handling and measurement of interactions.
There is only one problem. The next era of customer experience may not be about managing interactions more effectively. It may be about making many of them unnecessary. That distinction changes almost everything.
The dashboard can tell you what happened. The customer can tell you why.
One of the most significant implications of generative AI is not conversational automation at all. It is our newfound ability to listen.
For years, contact centres have sat on one of the richest sources of customer intelligence within the enterprise: millions of conversations that reveal what customers are struggling with, which processes are failing, where products are confusing, where promises are being broken, and where unnecessary effort is being created.
Yet we reduced much of that intelligence to categories, disposition codes and satisfaction scores. A CSAT score can tell us whether somebody was satisfied with an interaction. It cannot necessarily tell us whether the interaction should never have happened.
The distinction matters.
AI-enabled interaction analytics can now analyse large volumes of conversations, detect emerging patterns, interrogate customer intent and sentiment, identify root causes, and connect apparently isolated complaints to systemic problems.
What might appear as a handful of complaints, when conversations are analysed collectively, can reveal an operational problem that can then be addressed upstream.
That is not simply better reporting. It changes the purpose of customer operations.
Operational metrics do not disappear. They become hygiene factors rather than the strategic definition of success. The bigger prize is understanding and changing the conditions that created the demand.
Instead of asking, How did the contact centre perform yesterday?, leaders can increasingly ask: Why did these customers need us at all?
And eventually: What could we change so they don’t need us tomorrow?
AI can automate the old model. That doesn’t make it a new model.
This is where the current AI conversation turns uncomfortable. Much of the industry is deploying AI to speed up familiar tasks. Summarise the call. Assist the agent. Automate quality. Deflect the contact. Generate the response. Reduce after-call work.
Each can create genuine value. There is nothing wrong with optimisation. However, optimisation and transformation are not the same thing.
Organisations naturally use new technology to do yesterday’s work faster, cheaper and at greater scale. The greater opportunity is to use AI to do things the organisation could not do before.
That means the strategic question is no longer simply where AI fits within today’s operating model. It is whether today’s operating model survives AI.
Because a contact centre designed primarily around queues, channels, cases and labour capacity reflects an era when customers initiated service and organisations responded.
We are moving towards something different. Connected data, predictive analytics and AI can increasingly detect intent, identify risk, anticipate failure, recommend an intervention and, potentially, take action before the customer even enters a queue.
The emerging ambition is therefore not merely more personalised service but anticipatory service — moving from efficient reaction to intelligent intervention.
That does not mean predicting everything. Nor should every interaction be automated. The greatest value may lie in identifying the relatively small number of moments when timely intervention materially changes the customer’s outcome.
The operating philosophy shifts from contact handling to customer orchestration, raising an uncomfortable question for BPO. An economic contradiction sits beneath this transformation.
Imagine an outsourcing partner being paid primarily for people, seats, interactions or transactional activity. Now ask that partner to aggressively reduce interactions, automate tasks, redesign processes and reduce the human capacity required to perform them.
The more successful it is, the more effectively it can cannibalise its own revenue model. That is not fundamentally a technology problem. It is a commercial-model and incentive problem.
The traditional BPO model is unlikely to disappear. Human capability, specialist expertise, scalability and geographic delivery will remain highly important. However, the basis on which value is created — and increasingly contracted for — is likely to change.
The emerging opportunity is to move beyond selling capacity to helping clients continuously improve the economics and quality of customer operations.
That means the BPO of the future may increasingly resemble a transformation partner, a capability platform and a managed-services provider rather than simply an outsourced contact centre.
The winners will not necessarily be those with the most agents. They may be those that can orchestrate the most effective combination of human capability, AI, workflow, intelligence, and operational redesign.
The workforce does not disappear. Its purpose changes.
That also renders the simplistic debate about whether AI will replace agents increasingly unhelpful.
A growing share of routine work will be automated. Some roles will disappear or shrink; others will be redesigned, and entirely new combinations of human and machine work will emerge. Productivity expectations will shift.
But customer operations will still involve ambiguity, emotion, exceptions, vulnerability, commercial judgement and circumstances in which customers simply need another human being.
The more useful question, then, is not whether it is human or AI. It is which work should be performed by whom — and why? The emerging model is hybrid.
AI absorbs repetitive activity, searches for knowledge, analyses conversations, supports decision-making, and increasingly performs clearly bounded tasks.
Humans focus on judgement, complex problem-solving, relationship-building, handling exceptions, and the moments when trust matters most.
AI is almost like a new employee: give it a defined role, train it, monitor its performance, refine it, and gradually expand its responsibilities rather than attempting to transform the entire operation at once.
That is a useful antidote to another problem afflicting the industry: AI theatre. Leaders do not need another hundred AI demonstrations. They need evidence that a specific intervention improves an outcome.
Proof of value before promises of scale
This is why the next phase of CX transformation may also change the way technology is procured. The traditional enterprise technology journey often begins with platforms, procurement, implementation programmes and lengthy integration roadmaps.
AI creates another possibility. Identify a high-value operational problem. Find a suitable solution. Establish the data, governance and human controls needed to test it. Run a tightly bounded proof of value. Measure what actually changes. Then decide whether to scale.
The distinction between a proof of concept and a proof of value is important. The question is no longer merely whether the technology works. It is whether it delivers an outcome worth scaling.
Value should mean more than technical performance: customer outcome, cost-to-serve, employee impact, risk, revenue, or another measurable business result.
Transformation begins with the desired business vision, and people, processes, data and technology become interconnected building blocks for achieving it. Technology should therefore follow the problem, not the other way around.
A new gap is appearing between strategy and operations
However, a practical obstacle remains. Most organisations cannot continuously scan an exploding AI market, evaluate technologies, redesign customer journeys, rethink workforce models, fix processes, integrate data, manage implementation, operate the resulting environment and simultaneously run today’s contact centre.
This is where another category of provider becomes strategically interesting.
The next-generation managed service provider. Not the traditional MSP whose primary purpose is to operate technology. And not a strategy consultancy that disappears once it delivers the roadmap.
The emerging role sits at the intersection of strategy and execution. It starts with the customer and business outcome, works backwards into the operating model, and can then intervene across people, process and technology.
This does not outsource accountability for CX transformation.
Executive ownership must remain with the enterprise; the managed-services layer provides the capability, execution capacity and orchestration needed to translate that intent into sustained operational change.
Sometimes the answer will be process redesign. Sometimes it will be a workforce or capacity change. Sometimes it will be analytics. Sometimes it will be organisational capability. Sometimes an existing platform simply needs to be used properly.
And sometimes a high-potential AI solution deserves to be brought into the operation, tested against a real problem, and scaled if the economics and customer outcomes justify it.
The value isn’t owning every solution. It is orchestration.
That distinction could become increasingly important as enterprises seek to avoid replacing one generation of technology silos with an even larger collection of disconnected AI tools.
The evidence already points in this direction: sustainable AI transformation requires not merely technology but also ownership, governance, integrated platforms, operational expertise, knowledge management and continuous improvement.
Three futures are beginning to emerge
Look ahead several years, and at least three plausible futures for customer operations begin to emerge. They are not predictions. They are strategic choices organisations are already beginning to make.
In the first instance, organisations bolt AI onto the existing contact centre. Bots improve. Agents become more productive. Costs fall. But the architecture, incentives and operating philosophy remain largely intact. It is an AI-powered version of today’s model.
In the second, organisations pursue automation primarily as a labour-reduction strategy. Contacts disappear, but some of the understanding, judgement and human connection on which customer trust depends can disappear with them. Efficiency improves, while the experience quietly becomes more brittle. It is the automation trap.
The third potential future is more interesting. Customer conversations become an enterprise intelligence layer. Failure demand is systematically identified and reduced at source. Predictive signals trigger interventions before problems escalate. Routine activity flows between AI agents and automated workflows. Humans increasingly handle moments that require expertise, judgement and empathy. Commercial partners are rewarded for outcomes and continuous improvement, not the volume of work passing through the system.
Customer operations stop being the organisational safety net that catches everything the rest of the business gets wrong. It becomes one of the mechanisms through which the organisation continuously learns what to fix next.
That is not an upgraded contact centre. It is a fundamentally different operating model.
The real transformation starts upstream
Perhaps that is the paradox CX leaders now need to confront. For years, success meant becoming better at managing customer contacts. The next generation of success may mean deliberately removing some of them. Not all of them.
Some conversations build trust. Some reveal needs. Some create commercial opportunities. Some moments deserve — and will continue to deserve — extraordinary human service.
But avoidable demand is different. A customer contacting you because your process failed is not engagement. A repeat contact is not loyalty. A beautifully handled complaint remains a complaint.
An AI agent resolving a problem instantly is useful, but preventing the problem may be considerably more valuable.
The question leaders should be asking is no longer: How much of my contact centre can AI automate? It is: If we redesigned customer operations around outcomes rather than contacts, what would we build today?
Because the organisations that answer that question first will not simply operate more efficient contact centres; they may redefine what the contact centre is for.
