Trust First, Automate Second: Key Takeaways on Scaling AI in Customer Experience
June 4, 2026
Speed and cost savings. That’s usually where the AI conversation starts, and, too often, where it ends.
That was the thread running through our recent fireside chat on customer trust, where Five9 CTO and Head of AI Jonathan Rosenberg sat down with Rachel Munby, CMO of Anglian Home Improvements — the UK’s leading home improvement company, operating across a portfolio of brands including Anglian, SafeStyle and Everest.
What followed was one of the more honest conversations we’ve had about what it actually takes to scale contact center AI responsibly.
Here’s what stood out.
Personalization is the opportunity. Intrusiveness is the risk.
Anglian is a direct sales business with multiple human touchpoints — from call center agents to in-home surveyors — and for Rachel, the biggest opportunity for AI customer experience isn’t cost reduction. It’s personalization at scale, across the entire customer journey. “Right from the ad copy, right through to the website, and then ultimately into the call center.”
Rachel’s view is straightforward: if AI isn’t improving the customer experience, it shouldn’t be there. That sounds obvious, but it’s easy to lose sight of when the pressure to deploy is high and the use cases are multiplying.
The harder question is how far to take it. Personalization done well makes customers feel like a brand genuinely knows them. Taken too far, it tips into something that feels intrusive. That line is thin, and finding it — and staying on the right side of it — is one of the real challenges of scaling AI in customer experience today.
Be upfront with your customers or lose trust
By the same token, transparency isn’t a nice-to-have; it’s foundational. Anglian’s out-of-hours AI pilot is a great example of this in practice. Customers can book a sales appointment at 2 or 3 am, fully automated, with no human agent involved. But they know that going in. There’s no ambiguity about who — or what — they’re talking to.
Not every customer wants that. Some will always want a person. Others are happy to do it themselves, on their own schedule, without waiting on hold. The point isn’t to push everyone down the same path but to be honest about which path they’re on.
That’s because customer trust in AI breaks quickly when people feel misled. If someone thinks they’re talking to a person and later finds out it was AI, the issue is bigger than a bad interaction. It can damage the relationship entirely.
On the other hand, transparency alone isn’t enough. People need reassurance that the task was actually completed. If AI books an appointment, for example, send an immediate confirmation text directly from the booking system. Give customers something tangible they can trust.
Without that, many will still call back later just to make sure it worked.
Governance isn’t a compliance checkbox
Governance tends to get treated as a compliance exercise — something the legal team signs off on before you ship. That’s the wrong frame. Monitoring what the AI is doing, catching where it’s going wrong and understanding whether it’s actually accurate is what makes trust possible in the first place. It’s not a brake on innovation. It’s what keeps innovation from going off the rails.
The same logic applies internally. Contact center AI only works across a business when every function takes ownership of it — not just IT, not just marketing. It has to be treated as a capability the whole organization is building, not just a “tech upgrade” or a new tool being rolled out. As Rachel put it, “it’s a skill for everyone and a real strategic differentiator for the ones that get on board with it.”
Pick the low-hanging fruit first
The best place to start with contact center AI is wherever failure hurts the least. Post-call summarization is a good example of this. No customer ever sees it, agents get immediate value, and it’s fast to deploy. From there, move to simple self-service: appointment scheduling, cancellations, prescription refills. These generate high call volume, are relatively straightforward to automate, and give you something to build on.
Anglian’s out-of-hours pilot worked as a starting point for exactly this reason: it sat outside core business hours, which means there was room to learn without disrupting what was already working. The technology doesn’t have to be perfect on day one. It just has to be good enough to teach you something useful for day two. Or in Rachel’s words: “Don’t wait for the perfect start. Just get going.”
The contact centers that will get AI customer experience right aren’t necessarily the ones moving the fastest. They’re the ones being deliberate — clear about what they’re automating, honest with their customers, and willing to build trust before they build scale. As Rachel put it: “Don’t risk your core business processes. AI is supposed to complement and help you go faster — take friction out of the pain points, not rip up everything you’re doing today.”
