Customer service automation

Why we do not start with a chatbot

The chatbot is the most visible element of service automation — which is why it is often bought first. It is also the riskiest: it talks to customers unsupervised, and every invented or off-target answer spends trust that took years to build. Companies that have been burned by “automated” service were not burned by automation — they were burned by the wrong order.

Meanwhile, most of the time in customer service is not spent on the conversation itself. It goes on everything around it: identifying what the enquiry is about and who should handle it, finding the order and the customer’s history, checking the procedure, writing the reply, copying the outcome into your customer and sales management system (CRM). These stages can be automated with less risk, because a customer service adviser checks and approves the output before it is sent — and the customer gets an answer from a person. In the pilot, we check whether this reduces handling time.

What we actually automate

• Recognising and routing enquiries. Working out what a message from email, a form or chat is about — even when the customer used none of the expected words — assigning priority and directing it to the right person or queue. The most common first step: a mistake means the enquiry lands with another person in customer service who can already see the enquiries. Enquiries are only ever passed between people who already have access to them, never outside that group.
• Replies drafted for approval. The system prepares a draft answer from the knowledge base and the customer’s data; the adviser approves, corrects or rejects it. Every correction is also a data point about the system’s quality — the basis for later deciding which topics can move to unsupervised handling.
• A knowledge-base assistant for advisers. An answer to “what is the procedure in this case” with the source cited — instead of searching the internal wiki and asking on the team channel. It also shortens onboarding.
• Filling in the CRM and other systems. Copying the outcome of a conversation into system fields: topic, status, promises made to the customer. Work nobody enjoys, so it gets done late and incompletely — and it is what reporting rests on.
• Direct customer contact — at the end, not the beginning: on topics where the system has proved itself while a person approved every result, clearly labelled as automated, with the option to pass the matter to a person. Outside the team’s working hours, the system clearly says when and how an adviser will respond.

When enquiries require not just an answer but actions across several systems — checking stock, issuing a correction, changing an order — that is the line beyond which the work belongs to an AI agent. When an agent beats ordinary automation, and how its permissions are set, is covered on our AI agent implementation page.

The phone channel has rules of its own — speech recognition loses some of the words, and recording a call is a separate question about personal data. Which calls can be handed to a machine and which cannot is covered at voicebot implementation.

What we do not promise

• An unsupervised chatbot from day one. Everything that goes out to customers starts in a mode where an adviser approves it. Direct contact comes last — on topics where the system has proven its quality — and the decision to loosen oversight is made on pilot data, not by assumption.
• A percentage of automated enquiries up front. How much can be automated only becomes clear once you check what your real enquiries are about and count how often each topic comes up — and that is where we start. Stating a percentage before that count is an empty promise.
• A bot pretending to be human. When a customer is talking to a system, that is clearly labelled, and the customer can ask for the matter to be passed to an adviser. The full set of conditions under which we advise against a project is collected on what we don’t do.

Personal data in the enquiries

Customer service is, by definition, the processing of personal data — names, addresses, purchase history, sometimes health or financial details customers volunteer without being asked. Before any enquiry reaches an AI model, you have to settle which data leaves the company at all, where it goes and on what basis — from the choice of provider and processing region to limiting the data to the minimum a given step needs. How we approach it, and what to ask any supplier, is set out on our security and GDPR page.

How we work

We start by checking what your real enquiries from recent months were about and how often each topic came up. That shows what to automate first and what result can honestly be expected. Then one process and the smallest possible pilot, in which a person approves every result. Only the pilot results let us decide where checking can be reduced. The full course is set out in our implementation methodology, and the full scope of the service — from process audit through to upkeep — on our AI implementation page. The general conditions a process must meet to qualify for automation are set out on our AI automation page — customer service meets them more often than most processes in a business, because enquiries are text, they recur, and they leave a trail in systems.

What it costs

Cost is driven by the number of systems to connect, the state of the knowledge base, and how reliable the answers must be, that is, how rarely they may be wrong. The biggest hidden line is often the knowledge base itself: if the answers live in the advisers’ heads, writing them down is the first stage of the project — and it stays with the company whatever happens to the automation. All the components, upkeep included, are broken down in our note on what an AI implementation costs.

Frequently asked questions

Is customer service automation just a chatbot?

No — and confusing the two is expensive. A chatbot is the last, most visible and riskiest element: it talks to the customer unsupervised. Most of the value sits earlier and is invisible to the customer: identifying what each enquiry is about and routing it to the right person, drafting replies for a customer service adviser to approve, filling in the CRM, surfacing answers from the knowledge base. At these stages, an adviser checks and approves each draft before it is sent. This reduces the risk of incorrect information reaching customers, and in the pilot we measure answer quality and the time needed to review it.

What share of enquiries can be automated?

The honest answer: nobody knows until you check what your real enquiries are about and count how often each topic comes up — and that is where we start. In a typical support inbox, most enquiries concern a handful of topics: order status, returns, invoices, passwords. Those automate best. For the remaining, less common cases we automate the preparation of the answer rather than the answer itself. Stating a percentage before that count is an empty promise.

What about customers’ personal data in the conversations?

Customer enquiries are personal data, and sometimes special-category data too — when a customer volunteers health information, for example. So this has to be settled before the integration, not after it. You can control which data reaches the model, where it goes and on what legal basis: from the choice of provider and processing region to limiting the data to the minimum a given step needs. The principles, and the questions to ask any supplier, are set out on our security and GDPR page.

What does customer service automation cost?

Cost is driven by the number of systems to connect — mail, CRM, shop, order system — the state of the knowledge base, and how reliable the answers must be, that is, how rarely they may be wrong. A company with a tidy knowledge base and one CRM will pay less than one where the answers live in the advisers’ heads. Upkeep is a separate line: per-operation fees and keeping the knowledge base current, because the offer changes. We break the components down in our note on what an AI implementation costs.

Will customers not be put off by talking to a machine?

They are put off by a bad machine — one that does not understand the question, cannot hand the conversation to a person, or pretends to be one. That is why we recommend the reverse of the usual order: first automate the back office, where the customer still talks to a human, and only then — on topics where the system has proven its quality — direct contact, clearly labelled as automated, with the option to pass the matter to an adviser. Outside the team’s working hours, the system clearly says when and how an adviser will respond.