AI regulation is no longer just a tech or compliance issue, it’s becoming a boardroom priority.

In the US at a federal level the government seems to be actively opposed to AI regulation and in the UK, despite an interesting Private Member’s Bill, there’s no sign of any overarching AI law. But while the US and UK are still debating their approaches, the EU is ahead of the game with the world’s first comprehensive AI law: the EU AI Act. If you do business in or with Europe, this will affect you.

Why Should You Care?

No EU presence? Doesn’t matter. If you have EU customers or suppliers, you’ll likely be contractually required to meet the Act’s standards

Remember GDPR? The EU’s data privacy rules became the global benchmark. Expect the AI Act to have a similar impact

The Risk-Based Framework: What’s In, What’s Out

1. Unacceptable Risk: Banned

  • Social scoring, manipulative AI, and biometric categorisation based on sensitive traits are prohibited
  • Watch out: Using “black box” AI for things like fraud prevention or dynamic pricing could put you at risk

2. High-Risk AI: Strict Controls

  • Applies to recruitment, education, healthcare, credit scoring, policing, and safety-critical infrastructure
  • Requirements: Detailed risk assessments, transparency, human oversight, and conformity checks before launch
  • Don’t assume you’re exempt: Even apparently innocuous recruitment screening tools could be caught by these rules

3. General-Purpose & Generative AI: New Obligations

  • Foundation models (like ChatGPT or image generators) must ensure transparency, label AI-generated content, manage systemic risks, and clarify use of copyrighted data

4. Limited-Risk AI: Transparency Required

  • Chatbots and similar tools must clearly inform users they’re interacting with AI.
  • Heads up: Many bot providers still advise clients to hide from customers that they’re talking to machines —this will need to change

5. Minimal-Risk AI: Largely Unaffected

  • Spam filters, video game AI, and similar tools are mostly out of scope

The Compliance Challenge

For UK and global businesses, the message is clear: even without local laws, EU standards will shape your obligations. Cross-border operations will face growing compliance pressure, just as they did with GDPR.

Balancing Innovation and Compliance

The real challenge? Staying innovative while meeting new regulatory demands. Businesses must:

  • Identify which AI systems are in scope (which will include understanding exactly which parts of the business are using AI, to do what)
  • Ensure transparency and risk management
  • Be ready to demonstrate compliance to customers and partners

Need Help Navigating the EU AI Act?

At Customer Contact Panel, we help organisations find compliant, effective AI solutions—so you can innovate with confidence and accountability.

We’re caught between high expectations and uneven delivery. AI has the potential to transform contact centres, but only if implemented in a transparent, human-centred, and context-aware way. This article explores how consumer sentiment, operational strategy, and evolving AI technologies converge to reveal what works and what needs further improvement.

In our February 2025 whitepaper, Customer Contact Panel highlighted that this would be a ‘year of difficult conversations’ in which speed, automation, and empathy must be reconciled. AI can increase efficiency, but risks creating a sense of detachment if it isn’t matched with emotional intelligence. Interestingly, complementary research suggests people are more honest with AI when judgment is removed, particularly in sensitive domains such as mental health, financial support, or legal services. However, as the MaxContact report confirms, the majority of consumers still turn to voice when the stakes are high.

What Consumers Are Saying (And Why It Matters)

Voice AI by the Numbers — summarising adoption, customer preferences, and industry usage stats from the Synthflow whitepaper.

 

What does all this mean for CX leaders?

MaxContact’s survey found that 55% of people abandon calls due to long wait times, while 35% cite the agent’s lack of understanding. Complex account issues, payment negotiations, or emotional complaints are scenarios where empathy matters (and where automation often fails). The data reinforces what many CX leaders already sense: customers will accept AI for triage or routine tasks, but demand a human for anything nuanced.

Only 36% of respondents believe AI has improved their contact centre experience, and nearly 32% say it has made it worse. There’s a clear generational divide: 65% of 25-34 year-olds are comfortable with AI, but only 27% of over-55s feel the same. This generational lens is essential when planning AI and omnichannel strategies.

The core problem is bad AI, not AI itself. As noted in our earlier whitepaper, many AI deployments fail not due to technical limitations, but due to design and governance flaws. When AI is introduced without clear escalation paths, brand tone calibration, or decision traceability, customer confidence suffers. Mature solutions in the market now take a more human-aligned approach, creating AI agents that behave like brand-trained teammates, capable of recognising tone, understanding escalation logic, and respecting compliance frameworks.

Every decision should be traceable. Every transfer should carry context. These principles distinguish AI that scales from AI that stalls.

Omnichannel vs. Human-Centric: Getting the Balance Right

Consumers prefer voice support for immediate, emotionally resonant assistance. MaxContact’s research shows 60% view phone calls as the fastest route to resolution, far surpassing digital channels. Automation should manage repetitive tasks and noise, freeing up humans for high-value, high-empathy interactions. Smart triage, seamless handoffs, and transparent automation logic are crucial for omnichannel success.

Trust, Tone, and Transparency: Designing AI That Works

To address the most cited customer frustrations: poor escalation, limited response options, robotic tone – solutions must be designed with:

  • Cultural and tone calibration
  • Customisable escalation protocols
  • Transparent audit trails
  • Privacy-by-design aligned to GDPR and beyond

These aren’t technical ‘extras’, they are fundamental requirements in sectors where mistakes can harm trust, reputations, or wellbeing. In regulated or high-stakes categories such as healthcare, dating, or finance, the operational risk of misjudged automation is simply too high.

AI has advanced quickly, but trust remains fragile. Customers want efficiency, but not at the cost of clarity or empathy. The future of contact is digitally respectful, not just digital. The best AI solutions will pause, listen, and escalate when needed, not just answer fastest.

For contact centres navigating this balance in 2025, the opportunity lies in creating experiences that feel both seamless and human where AI takes the pressure off, but never takes over.

In our earlier whitepaper, we explored how AI adoption is reshaping customer contact – an area in which great risk and reward intersect. Six months on, the case for agentic AI has grown stronger, particularly in sensitive customer interactions where honesty and trust are essential.

Drawing on academic research and industry data, we now understand that AI can do more than just automate processes. It can unlock “more honest” conversations, especially in situations where fear of judgment by others or shame might inhibit disclosure.

This isn’t just a theory! Research from Stanford, MIT CSAIL, and NUS Business School reveals a striking trend: people are more open with AI than with human agents in contexts like mental health, financial distress, addiction, and relationship issues.

Why?

Because AI doesn’t judge.

Stanford calls this the social desirability bias, where people moderate their speech based on perceived perceptions. Removing this perception leads to greater honesty.

The ‘confession booth effect’, a term coined by NUS, also demonstrates this. In anonymised AI conversations, people admitted behaviours they hid from humans, like not reading terms and conditions or sharing passwords. In an insurance use case, initial disclosure accuracy rose by 40% when AI agents led the conversation.

MIT CSAIL found that people expend less mental energy managing impressions when talking to AI. This frees cognitive bandwidth for self-reflection and better problem-solving.

Now taking this approach, judgment-free AI agents can be implemented in high-trust, high-friction industries, such as mental health screening, legal triage, financial support, and trust & safety work. These artificial agents are more scalable and effective than humans.

The paradoxical truth is that people often feel more ‘heard’ by AI than by humans, because they don’t feel the need to pretend.

Yet traditional AI platforms struggle with emotional nuance, privacy, and secure escalation. It is essential to overcome these hurdles with strict compliance (GDPR+), contextual accuracy, and human-aligned escalation protocols.

The case for AI grows when combined with market data

Perhaps considered in the context of the long forecast demise of voice as a channel, recent research from the Synthflow white paper brings together a number of key usage stats which when considered with the findings of the academic research support the notion that AI voice will be here to stay?

 

What does all this mean for CX leaders?

  • Trust is key. Sensitive topics require the psychological safety of customers to be part your AI solution.
  • Design your AI agent around customer fears, not just FAQs.
  • Measure resolution accuracy and emotional sentiment, not just AHT.
  • Voice AI, when built correctly, can be the most honest channel for customers.
  • Integrated agentic AI ensures a consistent experience across platforms.

Customers don’t need AI to sound human. They need AI to “feel safe”. The leading approach to agentic AI will redefine what honest, efficient, and compliant customer interactions can be – especially in a world in which truth drives trust, and trust drives revenue.

Organisations that have adopted AI in their contact centres have often seen significant improvements, such as halved response times, 40–70% operational cost reductions, and increased contact handling capacity. However, as some partners have noted in their recent engagements with members of the the CCP team, these gains can be followed by a flattening curve and then a performance dip, if not implemented correctly.

We have revisited our February 2025 whitepaper, ’2025: A Year of Difficult Conversations’, in which we explored how AI, automation, and digital transformation would drive new operational and ethical challenges in customer contact. We previously highlighted the tension between cost optimisation, customer experience, and why thoughtful project governance will be required.

We thought it would be good to consider what may have changed and what lessons should be revisited.

Six months of continued observation and implementation across the market have revealed risks that automation without the appropriate planning and controls can have on your future operating model are more nuanced. While short-term AI gains are impressive, traditional approaches may erode long-term value through burnout, agent attrition, and customer dissatisfaction. This is the ‘AI Paradox’: the risk that productivity gains today may fuel tomorrow’s operational decline.

Beneath the surface, a gradual yet detrimental erosion of the human layer is occurring. Collaborating with AI often leads to front-line staff experiencing reduced recovery time, increased complexity in remaining ‘manual’ queries, and escalating customer expectations. Without adjustments to team structure, support, or metrics, burnout becomes a growing threat.

This productivity half-life, a period where efficiency peaks and subsequently declines due to human strain, is no longer merely a theoretical risk. Businesses are starting to witness this AI-driven degradation in tangible figures: within 18 months of implementing traditional AI, attrition rates rise by 65%, customer satisfaction scores decline by 20-30%, and agent engagement scores fall concurrently as the technology matures.

Agentic AI presents a more sustainable alternative. Instead of perceiving AI as a replacement for human input, CCP’s partners are illustrating how task-completing AI agents can alleviate the burden on agents, facilitate judgment-free conversations, and ensure capacity for the most significant human interactions when needed. Consequently, it not only yields improved outcomes for customers but also contributes to enhanced retention, reduced training expenses, and a more resilient workforce.

Mitigating the AI Burnout Trap: Lessons from the Last Six Months

  • Implement phased AI rollouts with human impact measures.
  • Adopt agentic AI that empowers humans, preserving judgment for complex cases.
  • Shift success metrics from AHT to FCR, CSAT, and agent engagement.
  • Involve agents in AI workflow design and iteration.
  • Regularly audit the AI-human balance: check whether tech amplifies or exhausts people?
  • Track attrition, training costs, and productivity when calculating your ROI.
  • Lead with transparency and ethics when deploying conversational automation.

Make certain you are on the right course

In short: if your AI roadmap doesn’t include agent wellbeing, then you’re building in risk. Efficiency must be sustainable, not just measurable.

Six months on, the market is beginning to learn this the hard way. The good news? There’s still time to course-correct. The AI paradox isn’t inevitable it’s just the result of decisions made without the full picture.

If you’d like to discuss in more detail how you can leverage the experience of our team and our partners, then feel free to contact us.

The health & wellbeing sector has always been rooted in human connection. Whether it’s supporting someone on their fitness journey, guiding a patient through treatment, or reassuring a customer about a sensitive health concern, the role of empathy is central.

But as the industry expands fuelled by digital-first healthtech, growing demand for wellness subscriptions, and rising consumer expectations, customer contact teams are under strain. The question for leaders is clear: how can we scale, stay compliant, and still deliver a deeply human experience?

From Cost Centre to Care Hub

Contact centres in health & wellbeing have traditionally been seen as a cost to control. Yet every conversation from a dietary query to a mental health support call has the potential to strengthen or weaken customer trust.

Forward-looking organisations are reframing service operations as a growth driver. For example, brands in this sector are exploring:
– Streamlined renewals and cancellations to reduce friction in subscription journeys.
– Smart routing for repeat callers, ensuring recurring issues are addressed quickly.
– Consistent omnichannel service so customers feel supported whether they call, chat, or message via an app.

When the contact centre is positioned as a core part of the brand experience, it moves beyond cost reduction and becomes a foundation for loyalty.

Automation with Empathy

The volume of routine contacts in this sector is significant – booking appointments, tracking deliveries, resetting passwords, updating payment details. These are tasks that can be handled by AI and digital workers, delivering instant, 24/7 responses.

The real opportunity is in blending automation with human empathy:
– Real-time agent assistance: AI surfaces the right knowledge at the right moment, helping advisors answer health or wellbeing queries accurately and sensitively.
– Vulnerability detection: AI can flag signs of distress in a caller’s tone or language, prompting the advisor to adapt their approach or escalate where appropriate.
– Conversation wrap-up & QA: Every interaction can be automatically summarised, with 100% of calls checked for compliance and quality, giving leaders confidence that standards are met consistently.

This partnership between people and technology doesn’t replace the human connection. It amplifies it, giving advisors the space to focus on empathy while automation handles repetitive, time-consuming tasks.

Scaling Securely & Sustainably

The growth in health & wellbeing services from digital fitness programmes to home diagnostics demands agile operating models. Customer demand can spike rapidly, whether during seasonal health peaks or major product launches.

To stay ahead, organisations are:
– Leveraging flexible sourcing models (nearshore, offshore, hybrid) to expand capacity quickly and cost-effectively.
– Adopting workforce management (WFM) tools to optimise scheduling and keep wait times short.
– Embedding compliance and security (PCI DSS, GDPR, sector-specific regulations) to ensure every interaction is safe and brand-protective.

By combining these capabilities, health & wellbeing brands can scale without losing sight of what matters most: trust, care, and the customer’s wellbeing journey.

The Strategic Shift Ahead

The contact centre is no longer just a helpdesk. It is becoming the front line of wellbeing experiences, where automation drives efficiency, and skilled advisors deliver empathy. For leaders in the sector, the challenge (and the opportunity) is to design operations that are both sustainable and human-centred.

At Customer Contact Panel, we connect organisations with over 220 delivery providers and 115 technology partners. We help health & wellbeing brands navigate their options, align technology with their customer journeys, and build resilient, customer-first operations fit for the decade ahead.

The promise of utilising AI in contact centres is enticing: faster service, lower costs, and increased productivity that pleases the CFO. Initially, the numbers support this claim. Efficiency increases by 30 to 55%, costs decrease, and there is the opportunity to scale without increasing headcount.

However, this success story hides a hidden threat that can quietly undermine every gain made.

It is the paradox at the heart of traditional AI implementations: the more you optimise for productivity, the faster you accelerate burnout.

If you’ve engaged them properly in your AI project, then at first agents will welcome the support. AI can take on routine admin, provide helpful prompts, and cut down on cognitive load, but as performance improves, expectations rise. The business begins to see these AI-powered gains as the new normal.

Over time, this creates a silent squeeze. Agents have less recovery time, more pressure, and fewer moments of meaningful human connection. Stress builds, job satisfaction falls, and staff turnover rises. It is suggested that within 12–24 months, many contact centres could face a reversal of their early gains.

This is the AI productivity half-life—a concept backed by both data and human psychology. Traditional AI shows impressive results in the first year, but by year two burnout sets in, staff turnover increases, training costs rise, CSAT drops and the contact centre enters a cycle of decline.

The numbers paint a sobering picture:

  • Training new agents costs £15,000–£25,000 per head
  • AI systems degrade when experienced agents leave
  • Customer loyalty falls when interactions lose their human touch

Ironically, the very systems designed to enhance performance can end up eroding it, because the human layer was never fully accounted for.

So, what is the alternative?

The better agentic AI solutions that we are now seeing take a fundamentally different approach. It’s not about replacing humans, it’s about augmenting them in sustainable, psychologically informed ways.

The AI agents don’t just answer questions; they complete tasks, make decisions, and manage workflows. They take pressure off humans without cutting them out.

And the results?

  • Better job satisfaction
  • Longer-lasting productivity
  • Reduced turnover
  • Stronger customer outcomes

More importantly, these systems are designed with long-term ROI in mind, not just how efficient your contact centre is today but how sustainable it will be two or three years from now.

How to Mitigate the AI Burnout Trap

  • Develop an agentic AI that reduces workload without diminishing human agency.
  • Shift KPIs beyond AHT to include FCR, CSAT, and engagement scores.
  • Implement phased rollouts to monitor human impact before scaling.
  • Prioritise agent training and involvement in AI design and deployment.
  • Track agent wellbeing alongside operational performance.
  • Combine automation with empowerment, ensuring humans retain control over complex and meaningful tasks.
  • Regularly audit AI value, evaluating cost, sustainability, and satisfaction.

Of all the contact centre use cases for AI, Pure Voice AI is the most disruptive – and potentially the most transformative. Unlike Agent Assist or auto-wrap that augment human performance, Pure Voice AI replaces the agent entirely for certain interactions.

What is the AI doing in Pure Voice AI?

Pure Voice AI uses fully autonomous AI agents capable of holding spoken conversations with customers—with no human agent in the conversation. For an inbound call, the AI could triage the call, and if it can deal with the interaction itself, it doesn’t need to trouble a human agent. If the enquiry does need a human agent, it can monitor who’s available and route the call to the next best available agent.

Ultimately, the idea is that these AI agents can answer questions, resolve issues, and even handle sensitive interactions such as payment disputes or appointment scheduling.

It’s far more sophisticated than IVR (interactive voice response) trees or chatbots. Pure Voice AI uses advanced natural language understanding, real-time decisioning, and speech synthesis to hold dynamic, human-like conversations.

Key benefits: 24/7 service

The benefits case here is far less about cost reductions, agent productivity gains and optimisations, as use cases 1-6 have already delivered well here.

It is far more about providing round the clock service and enhancing brand experience. Because we all have lives, and work, that mean calling between set hours can sometimes be difficult. But the reason many contact centres are not 24/7 with human agents is because the business case of the cost and overheads – from staff costs to heating and lighting – doesn’t stack up.

Smoothing demand

Not only is calling at set times difficult, it creates spikes in demand, for example around lunch time or just after work. What’s more, pro-active outbound calls can also be scheduled for more customer friendly times of day.

Multi-lingual cover

Where a contact centre needs to serve multiple languages, there is typically a primary language that most human agents speak, with a handful of specialists available for secondary languages. Which means that those secondary languages are a scarce resource, both in terms of availability and recruitment. With Pure Voice AI in the mix, it can detect the language being spoken and switch seamlessly into it.

Implementation considerations

While everyone is trying to rush to this use case, without computer use, proper integrations, optimised and redesigned processes, there is no real opportunity to leap-frog to full voice AI. Because the foundations are simply not in place to support it.

What can we expect to see?

While not quite there yet, it is just around the corner, and there will undoubtedly be a proliferation of pure voice AI, especially for outbound. Though businesses should expect regulation to swiftly follow.

As we await the true potential of Pure Voice AI, it is a case of charting a path to how you achieve this in future, not the focus for today. Down that road lies complexity, risk and far greater likelihood of project failure. When you could be realising value right now and incrementally from use cases that build the maturity on which to develop pure voice AI. A far safer path to value on every front.

To find out more about how CCP can help you make the right technology choices, read more here or get in touch.

This series of articles is drawn from our webinar with Jimmy Hosang, CEO and co-founder at Mojo CX. We explored seven key use cases for AI in contact centres, starting from the easiest productivity gains to value generating applications. You can find a summary of all seven use cases here, or watch the webinar in full here.

Agents often juggle multiple tasks during customer interactions, from information retrieval across systems, to research perhaps through search engines, and data entry to note-taking in CRMs and/or admin systems. Not to mention holding a conversation where they are listening and responding as naturally as possible. It’s a lot to ask while also ‘being present’ with the customer. The opportunity for errors and a sub-par conversation is obvious.

AI-driven hands-free conversations are designed to remove everything other than the conversation from the agent’s to do list.

What is the AI doing in Hands-free Conversations?

This again builds on the previous use cases as a good starting point. Think right back to use case 1 – autowrap, where the AI summarises call notes, and can either simply be copied and pasted into the CRM by the agent, or automated through deep integrations.

Imagine then a world, where not only does the AI do this, but it also navigates you though CRM screens as well as other platforms and apps, retrieving customer records and auto-populating information as you go. Meanwhile use case 5 – agent assist, is popping up with useful prompts to guide the call. Science fiction? Or science fact. The reality is that this is a genuine use case of today.

Key Benefits of Hands-Free Conversations: Absolute focus on the call

A truly liberating experience, the agent is focused solely on their conversation with the customer, while being fed the information they need to support the call and without worrying about what they are capturing as the AI is listening and interpreting to do that on their behalf.

The agent can listen intently, truly process the query and be mentally available to deliver responses where they’ve had the headspace to consider its appropriateness and the style of their delivery. Placing the human interaction at the very centre of the call to the exclusion of all other noise is extremely valuable when it comes to resolving that customer’s needs.

Reduced time spent on admin

If you were to say 10-20% of an agent’s time on a call is just typing and clicking to enter information and navigate screens, while a slightly arbitrary number, it’s inevitably slowing the call and reducing its value to the customer as the agent fills to give themselves the time to type.

Reduced keying errors

As the AI takes care of data entry, there are fewer agent keying errors. Not only does this reduce time on corrections, assuming there are field validations in place, or time taken to later interpret poorly captured data, it improves data quality overall. A key requirement for better analysis, better AI, better compliance, and better future performance.

Improved accessibility

What’s more, hands-free conversations can enhance accessibility – acting as a reasonable adjustment for people with visual impairments or limited hand function.

Implementation Considerations

First, the deep integrations necessary to support hands free integrations take time and shouldn’t be underestimated. Which is in part why this is use case 6 of 7. Because there will need to be some AI maturity building already to ensure both support for and success for this use case. But assuming you have that, it’s a natural progression to freeing agents simply to support customers.

However, there is another school of thought. Where the AI simply deals with all of the legacy for you. Which means you simply live with the poor processes, old mainframe systems, disparate add-ons and Excel spreadsheets you currently have, but without having to interact with them. The savings of not dealing with those, and not having to learn complex keying procedures to get to the screen you want, would be phenomenal and free cash for investment elsewhere. Listen from around 45 minutes into the webinar for Jimmy’s slightly mind-blowing hot take.

Second, accuracy and model training is paramount, which means training and testing the models will also take time and effort. As with other use cases, you will need to develop your own views of what is acceptable

Third, while it sounds all-encompassing, you could consider running the trained model locally and therefore reduce the computational costs.

Measuring Success

The primary KPIs here sit in customer satisfaction, agent productivity/average handling time and data entry or processing error rates. Beyond those, agent job satisfaction can be measured through feedback, attrition rates, etc.

But by far the most interesting benefit is the absolute focus on the customer and the delivery of superior service that should translate through to customer lifetime value.

To find out more about how CCP can help you make the right technology choices, read more here or get in touch.

This series of articles is drawn from our webinar with Jimmy Hosang, CEO and co-founder at Mojo CX. We explored seven key use cases for AI in contact centres, starting from the easiest productivity gains to value generating applications. You can find a summary of all seven use cases here, or watch the webinar in full here.

As self-serve improves, agents are increasingly left with the most complex queries.

Of course, it’s not always the case as many customers continue to prefer conversations over digital interactions, or are even digitally excluded, but for the most part, the expectation is that agents deal with call after call still, but with less and less respite where the query is nice and straightforward. And complex queries often involve multiple systems or multiple previous contacts, as they aim to follow the story. Which, if we go back to use case 1 – autowrap, we know our corporate memory can be patchy.

Yet customer expectation is that all of the necessary information is at the agent’s fingertips. And why shouldn’t they think that? Customers are often time poor and a bit frustrated already, so a longer than necessary call that may or may not answer their enquiry simply compounds their frustration.

What is the AI doing in Agent Assist?

As with all use cases to date, the AI is listening to calls to summarise key points. In Agent Assist – or ‘conversational guidance’, it is also retrieving and analysing previous data, so that, as with use cases such as auto coaching, it can make suggestions and guide processes during the customer interaction. It is, in effect, accessing and presenting the corporate memory that customers expect, but with bells on as it is also providing direction to the agent.

Key Benefits of Agent Assist: Assess customer needs

From the outset, the AI can help an agent understand what type of call they’re dealing with. Is this a customer with a quick query who is in a hurry, so it needs to be efficient, or do they have more complex needs that need to be addressed? Or even, is there an opportunity to cross- or up-sell to this customer?

Not only does this help to direct the nature of the call, it has an obvious impact on how the brand is perceived and even on topline revenue if it is possible to make a sale. Equally, it can prevent a clumsy attempt at a sale at the wrong moment, which could denude rather than enhance customer lifetime value.

This is also a key consideration in the push to self-serve, as customer experience becomes less personal and less represented by the people of the organisation. While a poor customer experience on a call can erode brand value, a great one can build far more than pure self-serve experiences.

Lower cognitive overload

In the context of agents needing to take more calls that are more complex, fatigue and cognitive overload is real. So while the last 15 years have been about focusing more on natural conversations and active listening, in a high-pressure, high-volume environment, doing that on every call is intense.

Of course, some agents may enjoy the need to think on their feet more than others, and therefore may be the ones who are trickier when it comes to adoption, but there are times when we all need a break from the mental strain of 100% concentration.

Optimise handling time

Agent assist can help to optimise handling time by keeping the agent on track, and reducing the amount of time spent unnecessarily building rapport. Of course some rapport is good, but if overdone, it can be confusing for the customer and result in repeat calls to resolve their actual query, rather than have a nice chat.

Accelerating the development of new starters

Dynamic conversational guidance allows new starters or less experienced agents to fly solo faster. They need to refer less to their supervisors for guidance (which in itself interrupts the flow of the call) and build confidence more quickly.

All of which translates to an increase in customer satisfaction from the use of corporate memory to deliver a better experience and faster call handling with responses that are right first time.

Implementation Considerations: Agent behaviour change

Aside from non-negotiables, such as good data (an absolute pre-requisite here) and systems integration, a key consideration for Agent Assist is to understand that it requires a much greater change in agent behaviour than the uses cases to date. Because you are in effect re-engineering conversations. So while it can accelerate the performance of a new starter, a longer tenured agent may be more ‘stuck in their ways’.

That means it typically takes 8 to 16 weeks to realise all the benefits of agent assist, possibly more if everyone is remote rather than office based. However, a two-to-four-month window to embed such change is both a really short time in the grand scheme of things and a small price to pay for the benefits available. Even considering that there will be a degree of things being a little slower in the first instance as you go through the J-curve of implementation.

However, almost all those aiming to realise AI benefits are jumping straight to this use case. Whereas they could prove the case for AI on much quicker and easier wins that are less likely to fail. What’s more, if those use cases have been implemented first, they lay the foundations for agent assist, and make both implementation and adoption easier and faster too. Without having proved the case for AI to the humans in the equation, adoption is slower, if accepted at all, and the benefits won’t come.

Knowledge base quality

Secondly, jumping too quickly to pure LLM Agent Assist and simply connecting to a knowledge base of historic conversations, then presenting information based on that alone as guidance, is a dangerous place to be. Because that assumes that all previous conversations were exactly as you wanted them, and not littered with the conversations of poorly trained or poor performing agents. Mojo’s advice here is to use the LLM to read in your script and build out the flow for that script and the associated dynamic pop ups. And it is essential that the knowledge base is part of continuous improvement too.

Measuring Success

There are some obvious key metrics that will be impacted by Agent Assist, from AHT and FCR to CSAT. More broadly, agent progression and job satisfaction can be measured through agent feedback, and customer lifetime value through customer analytics.

In summary, the key benefits are:

  • Increased agent satisfaction through reduced stress and increased performance
  • Increased customer satisfaction through faster and more accurate call handling
  • Reduced average handling time through more pointed conversations
  • Increased FCR through more accurate assessment and solutions
  • Improve opportunity spotting for x-sell and up-sell and guide to a sale
  • Improved customer lifetime value (CLV/LTV)

Done well, Agent Assist tools can enhance the capabilities of contact centre agents through intelligent support that enables agents to navigate complex queries or attempt to make sales with confidence. Which leads to more effective and satisfying customer interactions, greater customer lifetime value and even the potential to shift mindset from the contact centre as a cost centre to a value-driving profit centre.

To find out more about how CCP can help you make the right technology choices, read more here or get in touch.

This series of articles is drawn from our webinar with Jimmy Hosang, CEO and co-founder at Mojo CX. We explored seven key use cases for AI in contact centres, starting from the easiest productivity gains to value generating applications. You can find a summary of all seven use cases here, or watch the webinar in full here.

In an article published in the run up to Christmas 2024, I discussed the growing tendency in many organisations of the traditional Christmas peak starting to diminish, due to a variety of factors. Musing on the question of whether we had “…passed peak peak” (see what I did, there?) I wondered whether a reduction in the scale of ‘peak seasons’ was being paralleled by a growth in other sorts of both structural and spontaneous increases in contacts.

I was reminded of this article – and the discussions with colleagues and partners that inspired it – when I came across what was a new term to me, last week. The phrase is ‘perma-peak’ – the concept of a long-running or perpetual series of periods of high contact demand.

If the old, usually Christmas-related, peak season was the product of a society-wide, predictable (but hard to handle!) patterns of behaviour, then a perma-peak reflects a very different world. One in which both consumers and brands exhibit or can leverage greater personalisation, but in which demand patterns can spread and grow very rapidly:

1. Brands now have the ability to test and flex proposition and pricing, with instantaneous digital communication to customers and prospects
2. Social network effects can massively amplify the features and virtues of an offer or product – be that from an ill-prepared start-up or a settled, established big player. And, of course, negative aspects or faults can also be shared and multiplied, resulting in a flood of questions, complaints or claims

Pattern non-recognition

Despite advances in analytics automation and machine learning / AI, understanding and responding to demand changes in real time remains difficult and rare. Contact centres have always been operationally flexible and able to display great agility, but this is typically still a manual, even instinctive, response.

When insights are predominantly based on pattern recognition, increasing unpredictability makes operational flexibility a consistent challenge.

Likewise, agent assist and self-service tools and guidance will largely be optimised on the basis of prior experience. Unexpected demand spikes might be driven by either new query types and failure demand triggers, or factors in a novel combination.

Planning for the unplannable

In traditionally ‘peaky’ business sectors like retail, peak planning is an intrinsic part of how businesses are run. If you have a Christmas peak then thinking about how to handle it the following year typically starts in January, before the backlog of post-Christmas complex queries and complaints has even gone! Planning for the unpredictable is a different challenge.

So, if we really are entering an era of perma-peaks, will contract centre operations find themselves back in the era of watching service level charts drop precipitously whilst wondering what on earth is going on?

Not necessarily.

Perma-peaks need perma-flex

Just because tried and tested techniques are strained or start to fail, doesn’t mean that they can’t be revisited, adapted and fine-tuned.

• Workforce Management may become more orientated around rapidly identifying the different characteristics of peaks and corralling real-time shift flexing – and less about aligning scheduled staffing to forecast demand
• Your attempts to match staffing to demand may be less focused on seasonal efforts with shift banking and the use of temporary or fixed-term contract workers, but instead looking at incentivising intra-day shift flexibility or the peak use of a ‘gig’ model, using either in-house or outsourced workers
• Customer guidance, contact steering and self-service options will need to be able to be changed and weighted immediately in the light of peaks or troughs in demand. Root causes of demand spikes need to be identified and addressed – including restricting the promotion of or access to certain contact channels for limited periods, if required
• ‘Unforced errors’ by which organisations inadvertently trigger customer contact need to be avoided. Experience shows that the best way of doing this is through ensuring the contact centre has ‘a seat at the table’ when initiatives and campaigns are being designed

Whether you think that your next peak could be with you at any moment – or you’re confident that it will kick in from the 2nd week of October, just like every year – how to do you anticipate, plan and prepare?

Let us know; we’d love to share experiences and ideas.