By Debra Maxwell, CEO of ArvatoConnect
Customer service leaders are no longer asking whether automation belongs in the mix anymore. That debate is largely over. The harder part is far more practical: figuring out what to automate, what needs a human touch, and how to make the shift between the two feel completely natural for customers.
From a customer’s perspective, there should be no “channels” or “systems” at all. There is only the experience they have. And that experience either flows naturally or feels fractured.
If you’re trying to balance automation and human agents in your customer service operation without quality slipping, the answer starts simply but gets more demanding in execution.
I believe the answer lies in using automation and AI where they add the most value: handling predictable, repetitive and high-volume tasks, while also equipping human agents with better information, insight and guidance. Human agents should remain at the centre of emotional, complex or high-stakes interactions, for example, when a customer is talking about the financial difficulties of paying a bill.
The best handoffs between digital and human support are designed so well that customers don’t even notice the switch. This principle underpins much of the work we do; we’re passionate about designing automation around the customer journey, rather than bolting it on afterwards.
What and when can I automate?
Automation works best where the work is structured, repeatable and low in emotion: providing simple information about things like opening hours, or order tracking, returns, appointment scheduling, triage and routing. These are the highest-volume, lowest-stakes parts of the journey, and exactly where automation earns its keep.
It’s equally important to recognise that AI doesn’t just automate customer interactions; it can augment them too. Some of the most valuable use cases sit behind the scenes, helping agents identify signs of vulnerability, understand customer sentiment, access information more quickly and make more informed decisions. In these scenarios, technology enhances the quality of human interactions rather than replacing them.
This is also where the market is placing its bets. In our recent research, Care by Design, 36% of senior BFSI leaders said digital agents that guide customers through forms and signpost next steps had the greatest potential to support vulnerable customers, and 35% pointed to personalised journeys such as intelligent triage and specialist routing.
Handled well, this layer of automation removes traffic and noise, so agents can properly focus on interactions where judgement, reassurance and accountability matter. In ArvatoConnect implementations, we consistently see that when repetitive contact is reduced, agent capacity is reallocated into complex resolution and vulnerable customer support.
Keep humans where it counts
At ArvatoConnect, we’ve long held the view that roughly 70% of customer queries are routine, structured and well-suited to automation. The remaining 30% require a human, not because technology isn’t capable, but because some moments call for empathy, judgement and reassurance.
Customers tell us the same thing. More than half of financially vulnerable consumers surveyed said they would rather speak to a real person than AI when discussing complex or sensitive issues. A preference for self-service should never be assumed – and choice is important.
A customer explains their issue to a digital agent, whether that’s through chat, voice or another automated channel, gets passed to a human advisor, and then has to start again. I’ve watched that single moment of friction wipe out everything the technology was meant to achieve. The data backs up what I’m seeing: ArvatoConnect data found 35% of vulnerable consumers only reached a human after significant effort, and a further 15% couldn’t reach one at all.
In my experience, three things separate the firms that get this right:
- Pass over the full context: particularly with vulnerable customers, who so often have to repeat themselves.
- Make it easy to reach a human if they want to: no trap doors.
- Making confidence part of everyday culture
My key takeaways
The test I use with my own teams is simple, if it’s more routine and predictable, lean into automation. If it’s emotional or high stakes, keep it human. And if it escalates, make the handoff seamless – no exceptions.
I don’t believe quality, memorable customer service comes from choosing between technology and people. It comes from combining them well, so your agents spend their time on the moments that actually matter. And giving customers the choice.
And here’s what I’d say to any leader chasing the shiniest AI tool: the firms making real progress aren’t the ones with the most sophisticated AI. They’re the ones who’ve kept the fundamentals front and centre – customer service 101 – understand your customers, define what good outcome looks like, design around that, test properly, monitor continuously and adapt.
This post draws on findings from Care by Design: Making AI Work for Vulnerable Customers in Regulated Environments, ArvatoConnect’s survey of 1,000 senior decision-makers across UK banking, insurance, fintech, building societies and credit providers. Read the full report at arvatoconnect.co.uk
