Author transcript
Nicole Willing: The push for AI adoption is accelerating across the customer experience landscape. But for many CX leaders, the pressure to adopt AI is arriving faster than the organization’s ability to work out where it actually makes sense. Boards are asking teams to “do AI.” Vendors are promising major gains in productivity and customer experience. But on the front line, the reality can look quite different. So how should CX leaders navigate this gap? To discuss that, I’m joined today by Stephen Butler-Emmett, CX and AI Specialist at QContact. Welcome, Stephen. Thanks for joining us.
Stephen Butler-Emmett: Thank you for having me.
Nicole Willing: Great. So Stephen, let’s start with that tension between strategy and reality. Where is the biggest gap you’re seeing between what boards expect from AI and what teams are actually experiencing?
Stephen Butler-Emmett: What’s generally been coming out of board meetings is that the board expects AI to deflect half the contacts and cut costs overnight – that AI will be able to facilitate all our customer interactions and fulfill our containment strategy. The reality is quite different. Customer data is often very messy. You have frontline staff who are really stretched. And putting a customer-facing bot in place can often annoy customers more, and then the workforce has to handle even more hostile queries.
At the end of the day, customer queries are very nuanced. There are a lot of emotionally charged customers and a lot of sentiment involved. So boards are expecting these deflection strategies, but the landscape on the ground is quite different. I think the real wins right now are in the back office and the quality assurance space – birthing AI from the bottom up, and then moving into customer containment strategies with a much more nuanced and scientific approach based on the reality of what’s happening in your business.
Nicole Willing: Is it a question of expectations, or are organizations also underestimating the amount of operational work required to make these initiatives successful? Because it feels like sometimes they just think, “AI can do this, we’ll just throw AI at it.”
Stephen Butler-Emmett: The fundamental technology is brilliant, and it’s getting better almost daily. It’s really about doing the groundwork. The trap is that people want to rush in. They think AI can do everything without being guided and constructed in the right way, and rushing in can be super detrimental.
There are probably three areas where you can pay a price if you rush into AI. Customer trust is one. Customer-facing AI can be like a bouncer at a nightclub – if your bouncer isn’t well trained in dealing with guests, it can leave a very sour taste in a lot of customers’ mouths. Second, by the time that conversation reaches a human being, the customer can be incredibly hostile, which can definitely degrade agent morale and create a lot of skepticism around voice AI or agentic AI among the people doing the work. And third, boardroom credibility: if this doesn’t succeed and isn’t well implemented in your business, future AI initiatives can be challenging to get through at board level.
Nicole Willing: You touched on it there, but what are the real consequences for an organization if they push forward before they’re really ready?
Stephen Butler-Emmett: It’s exactly what we discussed. Firstly, it can be detrimental to your brand. Customers end up dealing with a brand that isn’t willing or able to deal with their queries, which makes them extremely frustrated. We’ve all dealt with those big companies that have these mechanisms, and it can be highly frustrating.
As we mentioned, it can also be detrimental to the workforce and really drive down morale. What can be extremely costly to a business isn’t only lower morale – it’s staff leaving. Highly competent staff who hold a lot of IP around dealing with these queries and with customers. And then there’s the credibility going forward: when new technology projects come up, the board can be a lot more skeptical. So it’s really in those three areas that it can be harmful.
Nicole Willing: Exactly. So this is where the operational side of CX is becoming so important. Stephen, you work closely with QA and performance teams. What does quality management look like day to day? Is the challenge collecting the data, or turning that data into action?
Stephen Butler-Emmett: QA is an interesting area, because the landscape has largely been unchanged for many years. Traditional QA means you take a number of interactions and have a quality assurance assessor go through them. They only have the capacity to review a handful of interactions per consultant per month. It’s extremely laborious and manual, and you end up with a tiny sample. You’re judging performance, and your understanding of what customers are saying, based on a tiny sample size. That can also be frustrating for consultants, because you’re looking at a handful of their interactions rather than a broad sample.
Essentially, traditional QA environments are scoring environments rather than coaching environments, and that’s where AI can lend a lot of support. It can go through a large number of interactions and give a much more objective view of what’s going on for a particular consultant. It can also be built as a coaching tool, because you don’t only flag the negative points – you can also highlight a lot of the great things a consultant is doing. That can be a really nice driver of motivation for staff, because they’re getting positive feedback about what they do well.
And with a broad sample rather than a very small one, you start to understand what customers are saying and why they’re interacting with you, and you can identify pain points much more quickly. We spoke about agentic AI – this is a brilliant way of birthing it. If you start identifying common queries that are fairly easy to handle from an AI perspective, that’s where you start getting the wins on the agentic side. It becomes a closed-loop solution if you do it well.
Nicole Willing: You mentioned some aspects that might be easy wins, but where should leaders be careful about moving too fast?
Stephen Butler-Emmett: Start where people are spending their time. Look at your workforce and understand what people are doing and what the common queries are. That’s a really good starting point. Then you can start creating augmented strategies. Certainly in this day and age, I wouldn’t leave everything up to AI. A really well-crafted augmented strategy is the way to go.
Practically, I’d say AI can probably handle about 50% of most companies’ queries quite well – the very repetitive, FAQ-type, structured queries. The other 50% still need to be dealt with by humans, where they’re a bit more emotionally charged. So start from the bottom up. Look at what customers are saying and how they’re interacting with you. Get the nuances from your quality assurance, understand the insights and the wrap-ups the consultants are doing, and build up from there. A good way to put it would be: augment first, automate second.
Nicole Willing: That makes sense. And to build on that, when it comes to automation, there’s obviously concern among agents about what it means for their jobs. How do managers make sure the message is that it’s supporting their development, rather than creating a culture of surveillance?
Stephen Butler-Emmett: Absolutely. We mentioned coaching earlier. AI in QA should really be framed as a coaching mechanism and tool within the business. It’s all well and good saying that – talk can be cheap – but when you start demonstrating the value it brings in making staff more successful, relieving them of tasks they don’t really want to do, and helping them deal with queries more effectively, that’s when you get internal buy-in.
Once they see it as a companion rather than a replacement, that’s really where you need to get to as a business. It’s about taking baby steps and showing them that it’s making them more successful and better at their jobs. Then it’s treated with far less skepticism than “this thing is here to replace me.”
If reducing human capital is part of the strategy, there’s always the natural attrition route, where automation can plug some gaps as staff naturally move on. But keeping a core team that’s highly capable of dealing with these queries, enabling them with AI tools to be more effective, and lifting the level of your customer experience through what your consultants do and how you handle queries – that’s a very powerful space to be in.
Nicole Willing: If we talk about expectations around ROI, what outcomes should a contact center realistically expect in the first six to 12 months?
Stephen Butler-Emmett: In the first three months, you’ll mostly be gleaning insights – configuring and iterating the solution and getting it to a place where it starts building IP around how your staff are performing tasks, what kinds of queries they’re handling, and what customers are telling you. So realistically, the first three months are about information gathering.
Between three and six months is where you start putting better solutions in place for how you deal with customers – potentially workflow and containment strategies, and better information available to consultants. That’s where it starts happening, and you can start measuring things like CSAT, first-call resolution, and whether sentiment in the queries is starting to lift.
Then between six and 12 months, you’ll start seeing the ROI in terms of how customers engage with you and, depending on the nature of your business, potentially more repeat business. That can be very measurable.
Nicole Willing: We’ve talked about board pressure, operational readiness and QA results. To bring it back to the people making these decisions: if you could offer one piece of practical advice to a leader who’s feeling the pressure to “do AI” but isn’t quite sure where to start, what would it be?
Stephen Butler-Emmett: Don’t start with “how do we do AI?” Start where the people are spending their time. As we mentioned earlier, understand what’s really going on in your customer interactions and build from there. Once you have a deep understanding of what’s happening at the coalface, what customers are saying, and the processes involved, start building on those insights.
You can take those insights back to board level and say: this is what the data is showing, this is where we believe the opportunity for AI is, and these are the queries that are very nuanced and still require a human being. That goes back to the iterative approach we spoke about – really taking the board on that iterative journey. This is what we’ve done, this is what we’ve found, and this is what we’re going to put in place.
The danger is that everyone gets sucked into the excitement and the pressure to deliver, and they automatically jump to “we’re going to create this containment strategy.” It’s not realistic, there’s a chance it falls over, and then the whole thing unravels from there. So start from the bottom up, understand what’s going on, and build an iterative journey that you can take customers on as well as the board. That would be my advice.
Nicole Willing: That’s really important advice. And for viewers who want to find out more about QContact and explore what you offer, where should they go?
Stephen Butler-Emmett: We’re on all the social platforms, so you can find us there as well as on our website. We’ve got some really good articles on AI and augmented AI, a whole range of AI features, and general omnichannel features as well. So please visit our website or find us on the socials.
Nicole Willing: Great. Thank you, Stephen, for joining us and sharing your perspective.
Stephen Butler-Emmett: Thank you very much for having me.
Nicole Willing: And thank you to our viewers. For more conversations and analysis on AI and the CX technology landscape, head to cxtoday.com. You can subscribe to our newsletter for the latest updates, and make sure to follow us on LinkedIn. Thanks for watching, and see you next time.
