Anyone who says "our work is full of customer contact" is not describing a single task, but dozens. A question about an outstanding invoice, a complaint about a delivery, a quote request, a renewal conversation, a crisis conversation with an angry customer: these are eight different things that happen to take the form of phone or chat. The fte scan from FTE TO AI therefore never puts "customer contact" as a whole on the scale. Every task is assessed separately, on eight axes, and with customer contact two of those axes weigh more heavily than with other work: how repeatable the conversation is, and what the costs are if it goes wrong.
Conversations that largely consist of retrieving and returning information shift the fastest. Order status, opening hours, an address change, a simple availability question: the task is essentially a search query with a human voice attached. Once the underlying systems return the correct data, there is little that an AI system is still missing here. You see the same movement in other departments: which work can ai take over in the finance department shows that checking invoices and sending payment reminders shift according to the same logic, namely because the answer is already in a system and only needs to be retrieved.
Conversations that repeat in fixed patterns, such as the initial handling of a complaint or asking questions about a fault, also shift considerably. Not because these are unimportant conversations, but because the pattern occurs often enough to model. What AI delivers here is an initial handling that goes to a human as soon as the pattern does not match.
Two categories of customer contact remain with humans, and that is exactly where much of the confusion about "AI in customer service" comes from.
The first is the conversation with an angry, sad or uncertain customer where the outcome does not fit into a script. Not because AI cannot formulate empathetic sentences, but because what is at stake in the conversation is: convincing someone, making an exception, saving a relationship. That requires a judgment that no one can lay down in rules in advance.
The second is the conversation where a mistake turns out to be costly: a wrong commitment about a delivery time, an unjustified exemption, wrong advice about a contract. Here what counts is not how often it goes wrong, but what it costs if it goes wrong once. This trade-off is central to what does ai take over in work with high error costs: the higher the damage of a single mistake, the longer a task remains under supervision, even if the vast majority of cases would go without a hitch.
This is the core of why customer contact is so difficult to generalize. Two companies with exactly the same job description "customer service" can end up completely differently in the scan, and that is rarely because of the conversation itself.
The first difference lies in the systems behind the conversation. An employee who has to log into four systems during a phone call, retype data and wait for a colleague, does different work than an employee for whom all information is in one screen. Where systems do not connect to each other, the task is often not "having the conversation" but "having the conversation while being the link between systems yourself", and the latter is much harder to automate. This situation is described on what does ai take over if your systems are not connected: only once the underlying data is accessible and reliable can a task really shift.
The second difference lies in the question of whether the work falls under external supervision. Customer contact in a regulated sector, for example where a duty of advice or a duty of care applies, has requirements for recording and accountability that have nothing to do with how well AI can conduct a conversation. This is worked out on what does ai take over in work that falls under supervision: here it is not the difficulty of the conversation that determines the pace, but the rules around demonstrability.
The third difference lies in staffing. Companies with fixed office hours and a stable team assess customer contact differently than companies with shift work, peak hours or many rotating temporary staff, where the problem is not so much the task itself as the constant re-training of people on that same task. This difference comes back on what does ai take over with shift work and rotating staffing.
If customer contact is a fixed part of a role, nothing about the way personnel are treated changes because of the fact that part of the tasks shift. Which consequences an organization attaches to that is up to the organization itself, and for decisions that affect employment, separate legal requirements apply that this page does not provide for. What the scan delivers is a description of tasks and of where AI is already involved today, with or without supervision, not a statement about what a company should do with that.
A related part of customer contact is the internal conversation about it: handover, escalation, team meetings about ongoing complaints. That too is a task in itself, worked out on can ai take over leading weekly team meetings, and that assessment again runs via different axes than the customer conversation itself.
To see which part of your own customer contact falls into the category "can be taken over", "partly with supervision" or "remains human work", an indication can be obtained via the free quickscan: twelve questions, without an account, resulting in an indication of which part of the hours in your profile can be taken over by AI today. The full work scan, which assesses tasks down to the level of eight axes and converts them to fte capacity, is still under construction. What is already available now is an initial picture, not the complete judgment.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.