State of the Contact Centre Research 2026

As with almost all industries, AI is dominating the agenda when it comes to transformation projects and operational priorities, and contact centres are no different. Ed Creasey describes in the podcast that "the hybrid model, as we call it, so a bot or an agent agent and a human agent is now mainstream", and the research backs that up. Fewer than 2% of organisations say they have no plans for a blended AI and human journey, 65% say customer-facing AI is the largest area of AI spend, and 50% believe customers now expect 24/7 service.

However, Ed and Adam make an important distinction. Being "AI-led" does not mean that AI is handling everything. It means AI is becoming the first point of entry, the first layer of triage, and the first place leaders look for productivity gains. That is very different from saying the end-to-end customer journey has been solved.

The findings also show just how widespread this shift has become, with 60% of organisations already operating with both human and AI agents, 60% using AI as the default entry point while keeping access to humans, and only 21% still keeping human agents fully accessible from the start. That paints a clear picture of a market moving towards AI-first service, even if the back-end experience is often still catching up.

Adam reinforces that point by explaining that many businesses want AI at the front of the journey to deal with the mundane, repeatable work, leaving human agents to handle the more complicated and sensitive contacts. That model is clearly becoming standard, but both he and Ed are equally clear that they have not yet seen a perfect implementation. The opportunity is real, but so is the gap between ambition and execution.

One of the clearest messages from the State of the Contact Centre research is that as AI becomes more capable, expectations are not falling for human agents, they are rising. The report shows that 60% of contact centres are increasing expectations for AI-handled interactions, while 68% are raising customer satisfaction targets for human agents. At the same time, the top two priorities for human agents are improving efficiency and improving accuracy and consistency, which shows that businesses expect people to handle harder work while still delivering better outcomes.

Ed goes into detail about this tension between expectation and reality, stating that the use of AI "changes the role". As AI takes on more of the simple, repetitive and transactional work, the calls and chats that reach human agents are more likely to be complex, emotionally charged, complaint-driven or harder to resolve first time.

Adam Clough builds on that by explaining that many businesses are now redefining the role of the agent around empathy and complexity. In simple terms, customers may be happy for AI to deal with a meter reading, a change of details or a routine account query, but when the issue is sensitive, frustrating or higher stakes, they still want a person involved. That means the human role becomes more valuable, but it also becomes more demanding.

This is where the research becomes especially useful, because it challenges some of the old assumptions around performance. If agents are now dealing with the contacts that AI could not resolve, then standard measures like CSAT need more careful interpretation. As Ed points out in the session, a complaints-heavy role may produce lower satisfaction scores even when the agent has handled the conversation well, because the customer may still be unhappy with the outcome.

The takeaway is that AI does not remove pressure from the human role by default. In many contact centres, it is concentrating more pressure into a smaller number of harder interactions, which means leaders need to rethink what good performance looks like in an AI-enabled environment.

The research makes it clear that contact centres are not moving towards a simple replacement model where AI takes over and humans step back. Instead, they are moving towards a more blended operating model, and that model is more complex to run than it may first appear. AI might absorb more transactional contacts, but the work of planning, forecasting, scheduling and balancing service still remains. In some cases, it becomes harder rather than easier.

Adam explains this particularly well from a workforce planning perspective. He argues that AI may increasingly handle the on-the-day, transactional work, while human teams focus on predictive, complex or emotionally charged activity. The challenge is that WFM teams still need to forecast those AI interactions, because most AI models are transactional and still carry cost. That means the planning team is no longer just forecasting human demand, it is forecasting the whole blended estate.

That is why Adam warns that "it [AI] will change how those teams have to operate and that world's going to be a lot more complicated". AI can automate some of the manual monitoring and recommendation work that sits within real-time and intraday teams, but that does not remove the need for human judgement. It changes where that judgement is applied and what kind of decisions planning teams are expected to make.

The takeaway is that good AI and human operating models are not simpler versions of the old model. They are more layered, more interdependent and more demanding, which means planning, operations and service design need to evolve with them.

One of the most useful findings in the research is also one of the most revealing. When customers move from AI to a human agent, only 53% of organisations say full conversation history is passed across channels. A further 33% say some history is passed but with limited context or only summaries, while 10% say only partial information is available and 3% are unsure. That means 44% of companies are still giving human agents only partial or no visi. That means 44% of companies are still giving human agents only partial or no visibility into prior AI interactions.

That gap shows up clearly in the podcast discussion. Ed repeatedly comes back to the idea that AI has to fit into the right workflow, because a good front-end interaction can still produce a bad customer experience if the handover is poor. His examples from fintech and white goods manufacturer make that point very well. In both cases, the bot experience looked promising, but when the journey needed memory, continuity, escalation or follow-up, the design started to fall apart.

Adam makes a similar point when he talks about organisations using AI as a point solution. Too often, businesses try to solve one visible part of the process without thinking through what happens next, what happens if the customer needs to go backwards in the journey, or how the human agent picks things up without having to piece everything together manually. That is where customer frustration tends to build, even if the AI itself appears to be working as intended.

The research presents this tension very clearly by describing AI for agents as "support in theory, but adding pressure in practice". That is a useful phrase, because it captures the reality many contact centres are now facing. AI can absolutely remove repetitive admin and routine contacts, but that does not automatically make the working day easier for human agents. In some environments, it may make it more intense.

Adam explains that the more efficient businesses make their AI, the less efficient they may have to allow their human agents to be. That is because the simple work disappears, average handle times can shorten in the wrong places, and agents can end up with more cognitively and emotionally demanding interactions back to back, with less natural breathing space between them.

Ed adds that leaders need to "have a lot of empathy for what AI is going to do to their workload", which is a helpful corrective to the more simplistic productivity narrative. If AI deployment is only judged by speed, cost or automation rates, then there is a real risk that the human role becomes harder without leaders fully recognising it. In that scenario, agents end up carrying more emotional load while the organisation assumes the technology has reduced pressure.

The research does not argue against AI, and neither does the podcast. What both suggest is that agent experience is now a design question. If businesses want AI to support people rather than exhaust them, they need to rethink occupancy, shrinkage, coaching time and the structure of the working day, not just the technology stack.

One of the strongest areas in the research is the growing role of AI in quality management, coaching and training. The report shows that 62% of contact centres now use Auto QM (Quality Management) for both AI and human agents, 53% use AI to help supervisors coach agents, 45% use it to generate written evaluations, and 38% use AI for automated training. Most notably, 75% report improved performance when AI supports training and evaluation.

Ed argues that one of the biggest benefits of AI in quality is that it moves people away from manual form-filling and towards actual coaching, stating that "instead of filling out forms, they're coaching". AI does not remove the need for leadership or judgement, but it can remove a lot of the repetitive effort that gets in the way of it.

Adam then takes that one step further by saying that "AI flips the analyst role from finding the problem to working out the solution". Whereas historically, low sample sizes and manual effort meant analysts and QA (Quality Assurance) teams could spend huge amounts of time just locating issues. With AI, much more of the problem discovery work can be done at scale, which means more time can be spent deciding what to do next.

The research also includes a strong supporting example from a UK QA manager, who said an AI evaluation was "far more comprehensive" than the team manager comments and more clearly picked up missed opportunities to de-escalate the call. That reinforces the wider point that consistency and scale can be major strengths of AI in quality, as long as the organisation still has the time, ownership and coaching discipline to act on what the insight shows.

Another fascinating finding in the research is that AI enablement has outpaced AI education. Only 35% of agents say they know which tools in their contact centre use AI, while 55% fear AI could change or replace their role. At the same time, 48% want more AI tools, and 60% of younger generations are already using generative AI in and outside work. The appetite is clearly there, but the understanding and communication around rollout are still inconsistent.

Ed makes this point directly in the discussion, saying that the next phase is not just about deploying AI, it is about explaining it to agents. That feels especially important because the podcast repeatedly shows that poor rollout does not just create nerves, it creates bad implementation outcomes. Oliver highlights the example of auto-summary being switched off because it was producing poor outputs that agents had to rewrite, which is exactly the kind of experience that damages trust early.

Adam's view is that agents should be much more involved earlier in the rollout process, not simply trained at the end of it. He argues that frontline teams will spot process holes faster than senior stakeholders because they understand where journeys fail, where systems frustrate the customer, and where real workarounds already exist.

The research gives helpful weight to that argument, because it shows that resistance is not the whole story. Many agents are not anti-AI at all. They are open to it, already using it, and can see where it may help. The problem is that too many organisations are still deploying faster than they are engaging.

One of the most significant shifts in the State of the Contact Centre research is how leaders now see the role of the contact centre itself. 61% say executive leaders view the contact centre as a value centre, and among those still seen as cost centres, 60% want to become profit or value centres. That marks a clear move away from the older view of the contact centre as simply a labour-heavy service function.

The reason for that shift is also set out clearly in the research. Conversation intelligence is making far more of the interaction estate visible and usable. Around 49.2% are already using conversation intelligence, while 97% of contact centres using Workforce Engagement Management (WEM) solutions plan to have implemented it by 2027. The insights being extracted are not being kept inside the contact centre either. The report shows they are being actively shared across executive leadership, sales, product development and marketing.

Ed explains that organisations are finally able to use contact centre insight to understand what is happening elsewhere in the business. Instead of just dealing with downstream demand, they can now identify which upstream decisions, product issues or service failures are creating unnecessary contact in the first place. Adam adds an important note here too, which is that value is not only monetary. Better visibility can create strategic value even when it does not show up as direct revenue.

The research also highlights that AI isn't only changing the role of the agent, but the team leader and supervisors as well. The report shows that 53% of supervisors use AI for interaction analysis, 50% for scheduling and forecasting, and 45% for coaching support. It also finds that 97% say AI has reduced repetitive manual workload to some degree, while 95% want more AI-powered tools to support supervisors.

Ed suggests in the session that team leaders may actually be more affected than agents in some areas, and that feels consistent with the research. Much of the management layer has historically been tied up in repetitive evaluation work, reporting, administration and low-sample review. These are exactly the kinds of tasks where AI can make an immediate difference, not by replacing leadership, but by removing manual work that gets in the way of it.

Adam makes an important distinction, though. AI can help with queue monitoring, recommendations, trend spotting and identifying coaching opportunities, but it cannot yet replace the human coaching role itself. It cannot read an individual in the same way a strong team leader can, nor can it adapt coaching style with the flexibility or judgement that people bring to those conversations. That means the role becomes less administrative and more human, not less important.

The State of the Contact Centre research is especially clear on where training priorities are heading. Empathy and emotional intelligence rank highest, followed by resilience and compliance, with troubleshooting and product knowledge coming after that. That alone tells an important story. As AI takes on more straightforward interactions, the capabilities that matter most for human agents are becoming more human.

Ed spends time in the podcast unpacking what empathy actually means, and his point is a useful one. Empathy is not simply sympathy, and it is not always about sounding softer or slower. It is about getting on the customer's wavelength. That might mean calming someone down, but it might also mean matching pace, urgency or tone in a way that helps the conversation move forward. For example, Ed explains that if a customer states that they are in a rush as they are catching a train, empathy would be understanding the situation and trying to get their request resolved hastily.

That matters because AI may be able to identify patterns in sentiment, escalation or failed interactions, but it does not yet remove the need for human interpretation. Adam makes that point well when he says that actionable insight is what really matters. Data may tell a leader that a certain behaviour is happening repeatedly, but it still takes judgement to turn that into the right kind of coaching or development for the person involved.

The discussion returns several times to a point that the research helps explain very well, which is that many organisations are still underestimating the effort needed to make AI work properly. The technology may be moving quickly, but that does not mean implementation has become simple. In fact, the more mature the ambition, the more important data structure, workflow design and operational ownership become.

Adam is especially clear on this when he says that AI is "not an off the shelf plug-in in 99% of cases". His point is not that the technology lacks value. It is that businesses often underestimate the amount of thought, tuning and process design needed to make it work in their own environment. Ed makes a similar point from a different angle by stressing the importance of domain expertise, metadata and understanding the difference between a good AI use case and a badly fitted one.

The research provides a helpful structural explanation for why this is so difficult. It identifies fragmented data as the number one issue stopping organisations moving from cost centre to value centre, and shows that 66% of organisations use three or more contact centre vendors. In practice, as the podcast suggests, many estates feel more fragmented than that. It is very hard to create joined-up AI experiences on top of disconnected systems and inconsistent structures.

That is why both Ed and Adam advocate a more realistic, use-case-led approach. Rather than trying to transform everything in one go, they talk about proving value in one process, learning from it, and then expanding into the next. That is not a lack of ambition. It is usually the only way to create sustainable AI change without overwhelming the business.

By the end of the session, the most useful definition of good AI implementation is not technical at all. It is operational. A well-run AI-enabled contact centre is one where self-service works when it should, context persists when customers move channels, agents get the right support at the right moment, and leaders have the visibility they need to make better decisions. That is a much more meaningful measure of maturity than simply counting how many AI tools have been deployed.

Ed describes this day-to-day model very clearly through a set of practical markers.

• "Seamless self-service."
• "Context persists."
• "I get real-time support."
• "I've got my data all in one place."

Those phrases work because they describe the experience of a well-run contact centre, not just the presence of technology. They also line up closely with the biggest findings in the research, particularly around hybrid service, handover quality, workforce support and the value of unified intelligence.

Adam then adds the workforce planning perspective, explaining that AI should enable faster real-time response and better operational adjustments across the day. That includes helping leaders respond more intelligently to demand, but also helping them create the space for coaching, support and better people management. Ed's example of remote-agent friction and system latency adds another useful layer, showing that AI and analytics can also expose hidden operational issues that were previously invisible.

Taken together, the research and the podcast point to the same conclusion. Good AI implementation is not defined by how much AI is present. It is defined by whether the contact centre feels more joined up, more informed, more supportive and easier to run because of it. That is what leaders should be assessing day to day.