AI by department · customer service
AI in customer service: the answer is ready before the ticket opens
For customer service teams that get hundreds of questions every day by email, chat and phone. AI reads the question, looks up the order and the policy and prepares an answer in your help desk, in the customer's language. Your employee checks and sends. The customer gets an answer sooner, and the team spends less time searching.
- 80% of 6,000 questions a month prepared
- Answers in the customer's language
- A human sends
Customer service agent
livereads 212 new tickets from email and chat
looks up orders in the ERP and the status with the carrier
labels all tickets and puts the complaints in the second-line queue
Worked with us
01 / What it is
What does AI do in customer service?
AI in customer service does the searching and typing behind an answer. It reads the question, looks up the order, the contract or the policy and prepares an answer in your help desk, in the customer's language. It also labels and routes tickets and keeps the knowledge base up to date. Your employee checks, adjusts where needed and sends.
That sounds like a chatbot, but it's something else. A chatbot sits in front of your customer service. We usually build behind your customer service: the AI works for your employee, not instead of them. And the answer remains your answer.
Why behind? Because most customer questions are just not quite standard. 'Where is my parcel?' is easy. 'Where is my parcel, and can I collect it, because I'm not home on Thursday?' less so. A wrong answer to that costs you a customer. A draft that an employee checks in ten seconds costs you nothing.
01Preparing answers
With the right order, the right policy and in your brand's tone of voice.
02Labelling and routing
Every ticket gets the right label, the right priority and the right queue.
03Multilingual
Answers in the customer's language. Your employee reads along in Dutch.
04Knowledge base
Questions without a good answer become visible. That way the knowledge base keeps growing.
02 / Sound familiar?
Where the time in customer service really goes.
Not to the conversation with the customer, but to everything around it: searching, copying, passing on and explaining again. At a few hundred tickets a day, that adds up to days a week.
The same question
“Most of what comes in today, we also answered last week. And the week before.”
Five screens open
“Help desk, ERP, the carrier's website, the customer file and yesterday's email. For one answer.”
Ask Piet
“Piet knows how the warranty works on old models. Piet works on Tuesdays and Thursdays.”
The wrong pile
“A complaint sat in the general inbox for three days before someone saw it had to go to second line.”
German customers
“One colleague writes good German. When they're on holiday, the German customers wait.”
After the peak
“After an outage there are six hundred tickets open. The rest of the week goes on catching up.”
03 / What AI takes over
Six places where AI takes over work in customer service.
From email to phone, and in all six a human stays in charge of what goes to the customer. We almost always start with the first card: that's where the volume is.
Email and tickets
Preparing answers
The question read, the order and the policy looked up, a draft in the help desk. The employee reads, adjusts and sends.
Routing
Labelling and routing tickets
Topic, urgency, language and tone determined as soon as the ticket comes in. Complaints to second line, invoices to finance, the rest into the right queue.
Languages
Multilingual customer service
The customer writes in German, your employee reads a summary in Dutch and approves a German reply. Without anyone needing to speak German.
Knowledge
A knowledge base that keeps growing
Answers come from your own knowledge base, with the source attached. Questions without a good answer become visible, so you know which article is missing.
Chat
Chat and email with the same knowledge
In the chat, AI can answer simple questions itself, with a quick handover to a human as soon as it gets tricky. The same knowledge, every channel.
Phone
Phone and voice
Calls summarized in the customer file, callback requests with the right context. A voice agent that answers the phone itself is also possible, but we rarely start there.
04 / Case
6,000 customer questions a month: from 30% to 80% with a prepared answer.
One of our enterprise clients gets around 6,000 customer questions a month. For every question, employees looked up the same order information and typed the same kind of answers over and over. For about 30 percent of the questions, something was already prepared. The rest started from a blank screen.
Now an agent prepares the answer before the ticket opens, with the right order information attached. For 80 percent of the questions there's a prepared answer when the employee opens the ticket. They check and send, always. Nothing goes to a customer without a human having seen it. Employees read and adjust instead of searching and typing. The time that frees up goes to the customer who really needs a human.
01Volume
6,000 customer questions a month.
02Before
30 percent of the questions with a prepared answer.
03After
80 percent, ready before the ticket opens.
04Human in the loop
An employee checks and sends every answer.
05 / Before and after
A day in customer service, before and after AI.
Six tasks as we build them in customer service. The time saved is an indication, except where a client figure is shown. What it delivers for you, we calculate with your own volumes.
- Enterprise client: 80% prepared
A standard question
Before
Open the ticket, find the order in the ERP, look up the status with the carrier, type an answer.
With AI
The answer is ready with the current status. If it's right, the employee sends it with one click.
A ticket comes in
Before
Someone reads the general inbox and drags tickets to the right folder. If there's time.
With AI
Label, priority, language and queue determined within a minute. A looming escalation goes to the top.
- Indication: no queue per language
A question in French
Before
Wait for the colleague who speaks French, or use a translation tool and hope it's right.
With AI
A summary in Dutch for the employee, a French reply to approve.
A tricky product question
Before
Ask around the department or search a folder of manuals. Ask Piet.
With AI
The answer from the knowledge base and the manual, with the source. Piet has time for his own work again.
- Indication: 2 to 4 min per conversation
After the conversation
Before
Retype notes into the customer file, often only at the end of the day.
With AI
Call or chat summarized, agreements and follow-up actions in the file. The employee checks it in ten seconds.
The weekly overview
Before
An export from the help desk, a pivot table and a guess at why it was busier.
With AI
Frequent questions, new topics and recurring complaints, with an explanation. Product and purchasing get a heads-up.
06 / Who does what?
What does AI do by itself, and where does a human stay?
Pick a task and see where it usually lands in the four roles of our APO method: human, AI assists, human approves or fully AI. In customer service, almost everything lands on 'human approves'. The conversation with an angry customer stays human work.
Pick a task
Human
A person does the work. AI stays out.
AI assists
A person does it, AI drafts or looks things up.
Human approves
AI does the work, a person checks before it goes out.
Fully AI
AI does it by itself. A person spot-checks.
Where it lands: Human approves
AI prepares the answer with the right order details. An employee reads along and sends: tone and exceptions stay human.
07 / Chatbot or not?
Should you put an AI chatbot on your customer service?
Sometimes yes, often not yet. An AI chatbot that answers by itself works well for questions with one clear answer that's in your systems: opening hours, order status, return conditions. For anything involving an exception, an emotion or money, you want a human.
That's why we usually turn it around. First we build the AI behind your team, preparing answers. Within a few weeks you see, per type of question, how often the draft is right without changes. What runs without errors for a long time can then move to the chat or to spot checks. What gets adjusted often stays with the human.
That way you build trust with evidence, not with hope. And you avoid the chatbot that sends customers round in circles until they call in anger. Still putting a chatbot directly in front of customers? Then make it clear that they're talking to AI, and give them a quick route to a human.
01Step 1: behind your team
AI prepares drafts, a human sends everything.
02Step 2: measure
Per type of question: how often is the draft right without changes?
03Step 3: let go of what's proven
Those questions move to the chat or to spot checks.
04Always: a route to a human
Anyone who wants to talk to a human gets a human.
08 / Work it out
What does AI deliver in your customer service?
Enter the number of customer questions per month and switch on what applies to you. The example deliberately calculates cautiously; in the Quick Scan we calculate with your own numbers.
AI drafts the answer, an employee checks it
Processes
€ 81,600saved per year
Above the € 80,000 threshold: this is where a partnership pays off.
€ 60 per hour · 4 min saved per question
09 / Yourself, a tool or a partner?
ChatGPT, a chatbot tool or an AI partner: what suits your customer service?
Three ways to get started. They aren't mutually exclusive, and for a small team the first or second is often enough.
| Yourself with ChatGPT or Copilot | Chatbot or AI feature in your help desk | AI Partnership with NinA | |
|---|---|---|---|
| What it does | An employee pastes in a question and gets a suggested text | Answers standard questions in the chat, summarizes tickets | Prepares answers with data from your systems, labels, routes and reports |
| Knows your orders and customers | · | Partly: what's in the help desk | ✓ |
| Works across systems (ERP, CRM, carrier) | · | Limited, depending on the integrations | ✓ |
| Answers in the customer's language | ✓ | ✓ | ✓ |
| A human approves before it goes to the customer | If the employee remembers to | Configurable, often not in the chat | Always, until it's proven per type of question |
| Who builds and maintains | Everyone for themselves | You, within what the tool allows | A dedicated team from NinA |
| Where the data is stored | Depends on the licence and agreements | With the tool's vendor | In the EU, with you or with a European provider |
| Suits | Small team, low volume, first steps | Many standard questions, one system | High volume, multiple systems, measurable gains |
Swipe sideways to see all columns →
Starting with the first column? A workshop gets your team to the same level in half a day. With the second column, we can build integrations alongside it. The third is our core work.
10 / How we approach it
From first conversation to AI in your help desk.
We build within the AI Partnership. You start with the € 6,000 AI Quick Scan: one day on site, and within a week you know what AI delivers in your customer service.
- 01
Intro call
Fifteen minutes
How many questions come in, through which channels, and in which systems is the answer?
- 02
AI Quick Scan
One day plus a week
We shadow your customer service and count, per type of question, the duration, the number of people and the frequency. You get a business case per opportunity and a twelve-month roadmap.
- 03
First process live
Quarter 1
Usually preparing answers for the high-volume question types. Connected to your help desk, tested on hundreds of real tickets from last month, with a human who sends everything.
- 04
Expand
After that
Labelling and routing, more languages, the knowledge base, the chat. Every month you see what it delivers in hours and euros, and what's next on the list.
11 / Cases
Customer service with AI, in practice.
From 6,000 customer questions a month to guest communication at Wens Chalets.
12 / Questions
Frequently asked questions about AI in customer service.
Can AI answer customer questions by itself, without a human?
Technically yes, but we always start with a draft answer that an employee checks and sends. At one of our enterprise clients, 80 percent of the 6,000 questions a month now have a prepared answer before the ticket opens, compared with 30 percent at first. Only when a type of question runs without errors for a long time can it move to the chat or to spot checks. Complaints, exceptions and emotion stay with a human.
What's the difference between an AI chatbot and an AI agent in customer service?
An AI chatbot sits in front of your customer service and talks directly to the customer. An AI agent works behind your customer service: it looks up information in multiple systems, prepares an answer, labels and routes tickets and presents the result to an employee. An agent can also handle a chat, but only for questions where the answer has been proven right. There's more explanation in what is an AI agent.
How do you automate customer service without it becoming impersonal?
By letting AI do the searching and people do the contact. The AI writes in your tone of voice, based on your own good answers, and an employee reads along before anything is sent. What customers mainly notice is that they get a good answer sooner.
Does AI in customer service also work in other languages?
Yes. The AI reads the question in the customer's language, gives your employee a summary in Dutch and prepares an answer in the customer's language. That way a Dutch team can help customers in German, French or English without a specialist per language. For sensitive answers, such as a complaint or a legal question, we agree that someone who knows the language reads along.
Which help desk and systems do you connect to?
To the help desk you already use, to email in Microsoft 365 or Google Workspace, to CRMs such as HubSpot, Salesforce and Microsoft Dynamics, to ERPs such as Exact, AFAS and SAP, to online stores such as Shopify and to carriers' systems. We usually build in n8n, with custom work where needed. The condition is API access or an export with the right permissions.
What does AI in customer service cost at NinA AI?
You start with the € 6,000 AI Quick Scan. Building happens within the AI Partnership: package S costs € 4,500 per month (30 hours), M € 7,500 (50 hours) and L € 12,000 (80 hours), for a minimum of twelve months. What goes live is covered by a maintenance contract from € 245 per month, including AI usage: Bronze runs on your own servers, Silver and Gold on our platform in the EU. All amounts excluding VAT. We only start if the partnership as a whole delivers at least twice the investment in the first year.
What about customer data and privacy?
Customer data stays in the EU, on your own infrastructure or on servers with a European provider. Azure OpenAI is an option. The AI only sees the data it needs for its task, everything is logged and the data agreements are set before we build. We work to ISO 27001 and NEN 7510; certification is in preparation. Code, data and documentation are yours.
Further reading
- AI in salesRequests in the CRM, quotes and follow-up that no longer falls through the cracks.
- AI agentsWhat an agent is, what it does and how we get it into production safely.
- AI PartnershipA dedicated AI team that builds, measures and maintains, process by process.
- Email automation with AIFrom writing assistant to an agent that processes email.
A customer service team that no longer falls behind?
Book an intro call. Together we'll look at your questions, channels and systems, and where AI saves time first.
01
Intro call
Fifteen minutes, no strings. Where does work get stuck and what do you want to achieve?
02
AI Quick Scan
One day on site. Within a week, a business case and a roadmap.
03
AI Partnership
A dedicated team that builds, trains and shows what it delivers every month.
160+ organizations · Sessions rated 9.3 · Based in Amsterdam