Artificial intelligence is rapidly becoming part of everyday business operations. From customer service and administration to data processing and decision support, organisations are increasingly using AI and automation to handle work that would previously have required human intervention.
The pace of that development is also generating debate. A recent BBC report highlighted concerns from former Anthropic researcher Jacob Coxon about the speed at which artificial intelligence is advancing and the potential risks associated with increasingly powerful systems.
The wider questions surrounding the future of AI are significant. But for businesses adopting the technology today, there is also a much more immediate operational question worth asking:
What happens when the AI doesn't work?
AI Creates Opportunity, and Dependency
There are compelling reasons for businesses to automate more of their operations. But as organisations become more reliant on AI, they also become more dependent on it.
Businesses already build resilience around other critical dependencies. They plan for system outages, telephony failures, cyber incidents, power disruption, loss of premises and shortages of key personnel.
As AI becomes embedded within business-critical processes, AI failure should increasingly form part of business continuity planning.
AI Doesn't Have to Collapse to Create a Business Continuity Problem
When we talk about AI failure, we don't necessarily mean the dramatic scenarios currently being debated by technologists and policymakers.
The risks facing businesses can be considerably more ordinary.
An AI platform could become unavailable. An integration with another system could fail. An automated process could encounter a situation it wasn't designed to handle. Responses could become unreliable or inappropriate. A security incident might require automated systems to be temporarily suspended.
Sometimes, the technology might be working perfectly well but the customer journey itself requires judgement, empathy or an understanding of context that means human intervention is still the better option.
The important question isn't whether technology can fail. It can.
The question is whether an organisation knows what happens next when it does.
Business Continuity Needs to Evolve With Automation
A good business continuity plan identifies the processes an organisation cannot afford to lose and establishes how those processes will continue when normal operations are disrupted.
Increasing automation doesn't change that principle. It simply changes some of the dependencies that need to be considered.
If an organisation moves a significant proportion of its customer service operation to AI, for example, its continuity planning should consider questions such as:
- Which customer journeys are dependent on AI?
- Which of those journeys are business-critical?
- What happens if the automated service becomes unavailable?
- Which interactions must immediately transfer to a human?
- What additional human capacity might be required during an outage?
- How quickly could that capacity be activated?
- What systems, processes and knowledge would those people require?
- How would the operation return to its normal model once the technology was restored?
Those questions become increasingly important as the proportion of an operation being automated increases.
The Human Layer Can Also Be the Resilience Layer
At Lemon, we don't believe the answer is to resist automation.
Technology should handle the interactions and processes it can perform effectively. But as businesses become more dependent on more and more complex levels of automation, human capability should also be considered as part of the resilience behind it.
Under normal operating conditions, AI might handle routine enquiries while a human team manages exceptions, complex interactions, sensitive situations or processes requiring judgement.
If the automated layer becomes unavailable, a predefined business continuity plan could temporarily expand the scope of the human operation, allowing priority customer journeys to continue until normal service is restored.
Rather than technology and people competing for the same work, they become different components of a more resilient operating model.
Planning the Human Fallback Before You Need It
The time to decide what happens when an automated process fails isn't during the failure. Human resilience needs to be designed into the operating model before it is needed.
Organisations introducing significant automation should consider their human fallback at the same time as they design the automated journey.
That means identifying which processes need continuity, documenting workflows, establishing escalation routes, determining the systems people would need to access and ensuring the required knowledge and capacity can be made available quickly.
An outsourced customer service partner can form part of that contingency model, using dedicated teams or shared services, to provide predefined human capacity without an organisation necessarily having to maintain the same level of contingency resource internally.
If a process matters enough to automate, it matters enough to consider what happens when that automation isn't available.
Automation and Human Support Should Be Designed Together
The organisations that benefit most from AI may not simply be those that automate the greatest proportion of their operations. They may be those that understand where automation works best, where people continue to add value, and how the two can support one another when something goes wrong.
At Lemon, we see an increasingly automated world creating a different role for human business process outsourcing, not competing with technology, but sitting alongside it when judgement, complexity, empathy or resilience requires a person.
Or, sometimes, quite simply when computer says no.
