An AI assistant might handle routine customer enquiries and support human agents handling interactions perfectly well on an ordinary Tuesday. But what happens when a power outage affects thousands of households; extreme weather disrupts transport networks, or a public safety incident sends call volumes soaring?
These are the moments when customer service automation faces its hardest test. Organizations must manage a sudden increase in demand while ensuring that customers with urgent problems can still get the right help quickly.
But a critical moment isn’t always defined by volume. For organizations responsible for critical national infrastructure, such as emergency services or utilities providers, any contact could relate to a safety risk or a vulnerable customer, whether it arrives in the middle of a storm or on a quiet night. In these environments, every interaction must be handled as critical, and the service must be just as dependable for one call as for 10,000.
Planning for these scenarios has always been part of the job of critical communications providers, and as AI becomes more deeply embedded in customer service, the same principle applies to automated interactions. Systems must be designed to withstand disruption without compromising the experience.
As Rhys Harris, AI Product Director, at Content Guru told CX Today:
“Your service needs to be like a glorious white swan—it needs to be serene above the water whilst the legs, your infrastructure layer, your application layer, your service layer, the design, are all doing the hard work and paddling away underneath the water.”
AI infrastructure may need additional compute capacity as demand rises, but the experience must remain dependable. “Ultimately, during these spikes in demand, the dials go up. You’ll be utilizing more compute, you’ll use more tokens, but the customer experience has to stay absolutely the same,” Harris said.
That requires more than meeting an availability target. A service can remain technically online while leaving customers waiting, delivering irrelevant answers or failing to recognize an emergency. Resilience must account for the person trying to get help, as well as the systems processing their request.
Not Every Interaction Should Be Treated the Same
When demand surges, treating every incoming contact with the same level of urgency risks overwhelming the service. Some customers need a routine update; others may be facing circumstances that require immediate human intervention.
AI can help absorb the high volume of straightforward enquiries, provided organizations establish clear boundaries around what should be automated.
“There’s going to be a large percentage of high volume, low intensity demand, and they’re often the best types of interaction to be able to automate,” Harris noted.
“That ultimately frees up human agents to be able to support complex, urgent, or emotional interactions.”
The challenge is identifying those interactions early enough. An AI agent needs to recognize when a customer’s circumstances indicate urgency and then route the conversation appropriately.
Organizations need to decide where the acceptable margin for error sits. In some circumstances, even a very small chance of an incorrect automated outcome may be unacceptable.
“If you can’t afford even a 0.1% chance of an AI-automated interaction getting the wrong outcome, you should either automate it through a deterministic pathway or capture some basic information and make sure it reaches a human who can handle it correctly. For critical infrastructure providers, that threshold applies to every interaction, not just those received during a surge.” Harris said.
AI should help determine the appropriate route for an interaction, rather than becoming a mandatory step every customer must navigate.
Proactive Communications Prevent the Queue Becoming the Customer Journey
One of the most effective ways to manage a demand surge is to prevent avoidable contacts from entering the queue in the first place.
Utilities offer a clear example, as operational monitoring systems and Internet of Things (IoT) devices can detect problems and trigger automatic updates in real time.
“By the time my power has gone off, and I’ve thought about contacting my energy provider, I’ve already got an SMS message to say, ‘we’re aware of an outage, and we’re on it,’” Harris said. “Not only is it typically a much lower cost to serve those contacts, but because it’s proactive, it’s a much better experience in a lot of cases as well.”