Author transcript
Contact Center Convergence: Closing the Gap Between AI Ambition and Operational Readiness
CX Today interview: Rob Wilkinson with Jonathan Kershaw, Director of Product Management, Vonage
Rob Wilkinson: Hello and welcome. I'm Rob Wilkinson and today we're taking a closer look at contact center technology in this new AI era. So, if you're challenged with rising service expectations, disconnected customer journeys, or the pressure to apply AI in a way that genuinely improves customer experience, then stay with us, because you'll get a practical takeaway that you can use this quarter.
I'm joined by Jonathan Kershaw, Director of Product Management at Vonage. He's an expert in this space. He's got a very clear view of the decisions, the risks, and the opportunities that leaders are facing today. So welcome, Jonathan. Thanks very much for joining me today.
Jonathan Kershaw: It's an absolute pleasure to be here. Thank you very much for inviting me, and I'm looking forward to a really exciting conversation.
Rob Wilkinson: Yes, it's really good to see you again. Before we get into the technology side of things, I'd just like, for our audience's context: what's changed for CX leaders over the past 12 months, especially with rising customer expectations?
Jonathan Kershaw: Such a simple question. But actually, yeah, a lot has changed, and I think we're probably seeing right now a lot about AI and trust changing, even as we're recording this. But to make it a bit more relevant, I think customers are now expecting instant, contextual and consistent service across every channel, and that's just got stronger and stronger.
So the bar has fundamentally shifted from "are agents available?" to "am I going to get an intelligent" – and we'll talk about this later – "hyper-personalized and contextual experience?" Dare I say it, people have now forgotten about COVID from that point of view. The tolerance for friction has gone. That's the real thing there.
Customers are comparing every voice channel to the best digital experiences that they've ever had, and the AI that people have been playing with in the back office has finally come to the front office, and people are expecting that. So customers increasingly expect that technology to know me – that classic "know your customer", which has been around for a while, but is now finally coming to life.
And CX leaders are being asked to deliver better outcomes. That's probably not new, but there's extra focus on it – with still tighter budgets and leaner teams. And ultimately, where this is leading is that because people are now seeing proof points from AI, you can no longer treat AI as just a future roadmap. Businesses do have successful implementations. So now it's a present-tense operational discussion rather than a future-tense one.
Rob Wilkinson: Nicely, succinctly put. That's very accurate, and I think anyone who's been around AI for any amount of time can see that we've definitely moved from those proofs of concept to it actually happening now.
So from the real vantage point that you have, speaking to organizations all the time who are dealing with these rollouts, where are they feeling the pressure most directly? Considering both the end customers who are trying to get help from them, but also the people on the front line who are trying to provide it.
Jonathan Kershaw: Yeah, I think that makes sense, and some of it actually overlaps a little. So for the customers, I think the biggest pain point is probably context loss. This isn't new. I reluctantly use the word "literally", but it was literally today: I was talking to a major railway company, I filled in the IVR, and then when I talked to somebody, they said, "And what's the answer to that question?" And I'm thinking, I've already done it. AI hasn't solved that. People are just putting AI in there.
So basically it's context loss – being transferred between channels or agents, and, in this day and age of hybrid workforces, from digital to human. Also, because of AI, voice is now becoming that high-stakes channel where, if a call goes wrong, it damages trust faster.
And that also actually bleeds into the business, because if a customer is unhappy – what was it, 1.7 unhappy experiences leads to churn – businesses are now running the risk of unhappy customers churning. So that's a big pressure.
And then agents was your other one. They're often being asked to handle more complex queries as AI absorbs the simpler ones. We know that, but they're doing it with fragmented tools: a bit of AI over there and a bit of voice over there – and we'll come to what we can do to prepare voice and AI later. All that gives you insufficient context as you go from digital to voice or whatever.
We all know agent experience is underinvested, because people are focusing on the customer journey. For a while there was a lot of talk about agent experience, and it really should resurge, because a good agent experience will drive a good customer experience.
And then, last but not least, that workforce pressure is real, and it's always been there. Nobody's saying it hasn't, but there's this whole "AI is going to replace me" school of thought. So there are the training gaps, the attrition that drives, and also the cognitive load of going from AI to there. But we know there's also the shift that actually AI will not replace – it will augment. That's one of those messages that you really need to curate and manage carefully, and that's where some of those huge friction points will also come from.
Rob Wilkinson: So yeah, that's interesting. The context loss we've always had: along the customer journey, there was always drop-off in terms of context. With AI came the opportunity to do it properly, but unfortunately things are still fragmented. It sounds like we're still having issues, because now it's between the different types of agents, whether that's the human or the AI agent, and the digital elements.
Okay. So when leaders say that they want an AI-first service model, what's your advice? What should they be careful not to lose along the way? Because that's got to be a risk, right?
Jonathan Kershaw: Yeah, and I think that's absolutely crucial – what not to lose. Again, it's the human touch. You have to trust who you're engaging with. And I think you made a good point: in the old ways, there was still a gap, a chasm, between the digital, the IVR, and the human. AI hasn't actually got rid of that. It's possibly exacerbated it, but that chasm is still there.
So, because of that, you need that human connection. We all know AI is going to handle the volume and the routine tasks, but the moments that matter most – that's where you don't want that chasm. They still require empathy, judgment, and accountability. That's what will make your high-value customers appreciate your customer experience over others.
We talked about context continuity – that transition. You cannot drop that. Yes, AI will solve simple things in isolation. That's great. But when that happy path doesn't work, is it smooth? People need to be considering that.
And then, I suppose, trust. We've talked about trust indirectly. You need good service. You need to know me. You need to have those proper guardrails. We've all heard stories where people have had their Java code validated [unclear] and they've then become racist. You have to have that trust. So build on that trust – it's just crucial.
And then actually, the people who get impacted the most – negatively if it goes wrong, but most positively if it's done well – are the employees. If we don't listen to their voice, those agents who are feeling bypassed rather than augmented are going to disengage. Your problems of poor morale, poor V [unclear], low staffing levels will just get worse.
So you need to have those agents engaged to have a successful deployment – to bring those models along, to test them, and to treat them as partners. In fact, we've had customers who've actually used their employees as pseudo customers. They've tested it, gone through their scenarios, and they know what customers or consumers are doing. So it's a really valuable way to test whether this is going to work. If it passes the employee test, there's a damn good chance it's going to pass the customer test.
And then, I guess lastly, measurement. With any change, there has to be measurement. So don't just optimize purely for cost or deflection. Are you changing the dial for the customer? Are they getting resolution, not just deflection? You'll hear the language change from deflection to resolution – "charged by resolution, not by interaction", those sorts of things.
But people also need to be aware that key metrics may change. AHT may actually go up for the human agent. But when you consider a hybrid workforce, overall it'll come down, because the AI agent's work might be shorter. You just need to consider those sorts of things. So, an element of ROI and proof that what you're doing is actually working.
Rob Wilkinson: It certainly confirms that we've seen the technology move so fast. Twelve months ago there were teething problems, POC issues, companies moving forward and making mistakes, and some of the horror stories we heard back then were more about people running before they could walk. Now this is really happening. The things we're hearing today from you are stories of people implementing things, and ways of avoiding issues when implementing. So it really sounds like, as you said at the beginning, the move has been made. It's not on the roadmap anymore. It is what it is.
So from that perspective, what are you seeing as the first readiness questions that a more cautious CX leader, who hasn't gone down this route yet, should be asking before moving from interest to actual implementation?
Jonathan Kershaw: Yeah. I'm going to say what everyone says, so this isn't insightful, but hopefully it sticks. It's data, data, data, data. Is your CRM clean? You might say, "Well, of course it's clean." But earlier on I talked about that classic know-your-customer personalization. If it doesn't know who I am, and it doesn't know my last interaction or the last update – because maybe you're only doing a 24-hour update across systems – you're giving a broken experience.
So is your data accessible and connected to the channels where customers interact? Also, if you're going to use AI at the front end, especially on voice channels, is your data accessible quickly? On a web chat, dead air, for want of a term, is acceptable. On voice, it is not. So it's not only whether your data is correct, but how fast you can get to it.
Also the processes: have you mapped those processes and key workflows clearly enough? You might say you have, but you've now been given an opportunity to change them. Are your processes there just because they've always been there? Now you've got AI in the mix, maybe it's a chance to change them.
And if you start to introduce AI and automation at scale, that is going to identify where you got it wrong – massively. Which takes us back to the agent experience: they are the ones who are going to have to mop it up. So think about those gaps.
You also alluded earlier to the fact that voice is different from the digital channels. We all think of it as one of our standard channels – and yes, 20 years ago it was, because it was the only channel. We talked earlier about digital giving the better experience, and we want voice to be the same. But voice doesn't do that, because you don't have the transcription. You need the transcription, and you need high-quality audio under that to deliver high-quality transcription, because it is that which will then drive all your AI tools on top of that. It sounds like a basic thing, but it's not even basic – it is fundamental if you want to take advantage of AI.
And by thinking about that, you're breaking that silo. Once you've got voice digitized with transcription, like you do for digital, you can use the same AI tools, the same insight, the same answers across all channels, the same personalized experience, and it becomes joined up.
And then governance. We talked about AI policies: who owns that escalation path? What happens if it goes wrong? I'm being quite quick on that, but I cannot emphasize it enough.
And sort of the last two, really. Do your agents have the tools and training they need to operate it effectively? Like all businesses, we've used AI ourselves, and you have to learn it and make the most of it. But you have to make sure that the metrics you're using to track adoption are good ones. You shouldn't just be asking, "Have you used AI?" "Yes, because I used AI to summarize my calendar for the day." It's: have you used it in a meaningful way that has actually improved your job?
And then lastly, what does success look like? I talked about it earlier – AHT changes. How do we know the model is working or not? For example, we offer automated disposition codes, as a simple one. It just saves the agent some time. But how often is the agent changing that? Is the AI actually working? Yes, you can deploy it, but are you getting the value? So you need to measure those sorts of things to prove success and refine it. And I'm not talking about tweaking the model, because that's just overly complicated. You want to tweak the implementation using the tools and the configuration you've got. The model should be abstracted away. So yes: measurement – is it working – with those examples.
Rob Wilkinson: That's a really comprehensive list, and I think it's super accurate as well. I like the fact that you've included the processes and the people in that list, because I think operations teams seem to forget, when they roll out new technology, that the processes are going to have to adapt. Even the operating model might have to completely change, and the training for people needs to reflect the new world. Quite often they get overlooked, so thanks so much for making sure they got called out.
And thinking about that, you've got a really great vantage point to give us some insights into the areas where organizations are likely to underestimate the work involved in connecting this AI service model with the reality of voice, context – to your earlier point – and everyday workflows. Where are organizations getting that wrong?
Jonathan Kershaw: [laughs] Again, a simple but direct-to-the-point question. I alluded to this earlier: people were putting AI on the voice channel and it was taking three or four seconds to get the data back – let's assume the data is correct. So it's things like that. Just because you've got AI, just because you've got voice, just because you've got CRM doesn't mean it's actually going to work.
So probably the first one would be CRM connectivity, but it comes down to data quality issues. We talked about whether you can get to that information. We all know that in a contact center, you've got one role, I've got another role, based on who we're talking to. Is that being reflected in the AI model? And are there legacy constraints that slow everything down? As I said, dead air is just a terrible thing, and we need to consider customer experience. None of this is in isolation.
And then again, context transfer and handoff. You might have invested a lot of money, time, and effort in the AI to handle those interactions, which is great. But if the handoff is slow and clunky – and I referred to one earlier, albeit in an IVR, where we've had years to get it better – if that handoff from AI to human isn't going to work, the ROI that you've built your model on – reduced AHT, more capacity – is gone. So you need to ensure that context is there at the right moment. The orchestration of that is detailed.
And also, as you hand those things off, there's the messiness of real-world workflows. That ties back to what we were talking about earlier: are your processes actually what you think they are? Because if the AI gets it wrong and kills the call, you're paying a price.
And then lastly, change management. Everyone knows this – we talked about it earlier. It's always people who are under-planned, but who actually underpin the success. Getting agents, supervisors, and IT teams aligned and confident takes time. I mentioned one of our customers earlier: they got their employees to test the scenarios, so they understood what the customer was going through and what was about to happen. So when the calls came in on escalation, they were ready. They'd, for want of a term, lived it [unclear]. They were prepared, bought into it, and understood it.
So there are some of those key areas that people underestimate – that connection. You actually have to consider these things. So, good question, but those are the areas.
Rob Wilkinson: I love that idea of bringing the employees in. It's almost like customer zero. Use your employees for that and you can't go wrong. You've covered an awful lot, Jonathan – thank you so much. You've given us a really good indication and removed some of the noise that's in this space.
What I'd like you to do, though – because there's quite a chunk of stuff there for people to take away – is this: if a CX leader has to pick one priority for the next six months, out of everything we've talked about today, or maybe something different, what would that be? What do you think they should focus on?
Jonathan Kershaw: Wow. So, one priority. I think the reality is that people will have high expectations for AI. We hear all the statistics – was it 85% have tried and 5% are successful? There's a reason for that. So basically, I guess: close that gap between AI ambition and operational readiness. I talked about something as simple as how fast the data is getting there. Even if you're preparing those bits, that's going to improve things for your normal agents handling calls. So align your ambition with reality.
I guess the ones that will be successful aren't the ones with the boldest AI visions, but the ones that have done the most foundational work to make it executable and differentiating for their customers. Which basically means – and I talked about it earlier – auditing where voice, CRM, data, and AI capabilities currently sit. If they are siloed, start to bring them together, but a little bit at a time. As I said: can you get the data to the human agent faster? Then it will be faster for the digital agent too.
We all want it to happen, and I know you said one, but there are lots of considerations, because everyone's different. Don't focus on the perfect platform moment, but one value workflow. Prove it. That's the thing, because there's so much noise, and nobody likes a failed project. People are now finally saying it has to be ROI. It's got to be measurable, successful, and actionable. So go with one that works, do that, and then build it out. Build your faith.
We said earlier: bring human agents into the process. What would they do? Where would they do it? They're living the business day in, day out.
And then lastly, partner strategically. There's a lot of choice out there, and no organization builds this alone. That classic phrase: no one person can build a toaster. Very simple, but no one person can do it. So understand what capabilities are in the platform – and by platform I mean CCaaS, AI, CRM – and which belong in the ecosystem as best of breed, and which are going to be critically strategic for the next six months and beyond. Don't just rip things out because you can, but validate that those different bits are giving you the value and quality you need for the foundations of the future.
So that might be five, but hopefully one should resonate with different people.
Rob Wilkinson: I think that's a fair point. [laughs] I had to try and see if we could get more. For me, the takeaway is quite clear: CX leaders need to keep the customer and the employee experience in view, understand where their data and workflows are today, and take a measured approach to bringing AI, voice, and CRM together in one place.
Unfortunately, that is all we've got time for. As always, I could talk to you all day about this, Jonathan. But before we go, if anyone in our audience wants to learn more or reach out to your team, what's the best way for them to do that?
Jonathan Kershaw: I think just the classic: go on to the website, have a search – there's a contact point there. Or go on to the various app exchange-type marketplaces and see people's comments. That would be the best and first way to do it.
Rob Wilkinson: Perfect. And you can also find a wealth of resources, related stories, and more videos just like this one at cxtoday.com. But for now, I've been Rob Wilkinson. Thank you very much for watching.