Skip to content
Solution · 01

AI agents for customer service and back office

An AI agent is software that carries out a task on its own, inside rules you set: it reads an incoming request, finds the answer in your own material and closes it — or passes it to a colleague with a draft attached. Goudbeek builds AI agents for businesses — for customer service and the back office — from Almere in the Netherlands.

Most of what fills a support inbox is the same twenty questions in different words. An agent can close those and leave the ones that need judgement on a colleague's desk, already drafted.

Who is an AI agent for?

  • Support teams whose queue is longer at the end of the day than at the start
  • Back offices where the same request gets typed out again every week
  • Anyone whose senior people are the bottleneck for routine answers

What does an AI agent actually do?

  1. 01

    It reads the whole request, not the subject line

    The ticket, the thread it belongs to, the order behind it and the customer's history. The same context a colleague would open before replying.

  2. 02

    It answers from your material

    Your help pages, policies, price list and order data — not from what a model happens to remember. When the answer is not in there, it is built to say so and pass the question on, instead of making one up.

  3. 03

    It knows which ones are not its to close

    Refunds over an amount you set, anything about a complaint, anything it is not confident on: those go to a person with a draft attached, so the handover costs a read rather than a restart.

  4. 04

    It leaves a trail

    Every action it took and the source it used, in a log you can read. When something goes wrong you can see exactly where, which is the difference between fixing it and guessing.

What may an AI agent never do alone?

  • Which systems it may read, and which it may only write to
  • What it may send without a human, down to the amount and the topic
  • The tone it replies in, and the answers it is never allowed to give
  • The off switch, in your hands, with no call to us required

What happens after go-live, and what does it cost?

We start on one real queue with your own data, so you can see whether it holds up before you commit to more. From there it runs on managed hosting from €39 a month, and the rules stay yours to change.

Read on in the knowledge base

Frequently asked questions

What is an AI agent?

An AI agent is software that carries out a defined task on its own. It reads an incoming request, finds the answer in your own material and closes it, or passes it to a colleague with a draft attached. What it may read, what it may send and what it may never do alone is written down before anything is built. Goudbeek builds AI agents for customer service and back offices. A longer explanation: what is an AI agent?

What is the difference between an AI agent and a chatbot?

An AI chatbot holds a conversation with a visitor on your website and answers from your catalogue and terms. An AI agent works a queue behind the scenes: tickets, mail or requests. Inside the limits you set, it can take steps in your systems, such as looking up an order or preparing a draft for a colleague. Both are built to answer only from your own material.

Which questions does an AI agent close on its own?

An AI agent closes the recurring questions whose answer is in your own material, such as where an order is, a change of address or a question about an invoice. You decide which topics it may close by itself. Refunds over an amount you choose, anything that smells of a complaint and anything it is not sure of go to a colleague, with a draft attached.

Where does the agent get its answers from?

An AI agent gets its answers from your own material: help pages, policies, price list and order data. Not from whatever a language model happens to remember. When the answer is not in there, it is built to say so and pass the question on, instead of making one up. Every step it took and the source it used go into a log you can read back, so when something goes wrong you can see where.

What may an AI agent never do on its own?

What an AI agent may never do on its own is yours to decide, and it is written down before anything is built. A common line: reading, searching and drafting are fine, but sending, paying, deleting and changing records wait for a person until you have watched it run long enough. Why that line sits at consequence rather than difficulty is explained in what an AI agent is allowed to do.

Can I switch the AI agent off or change it myself?

Yes, you can switch the AI agent off yourself: the off switch is in your hands, with no call to us required. The rules stay yours too: which systems it reads, what it may send without a person and the tone it answers in. They are written down before anything is built. Within that setup you can change them afterwards without the system having to be rebuilt; connecting an extra system is new work, agreed separately.

What does an AI agent cost?

The price of an AI agent depends on the scope: how many kinds of question it handles, which systems it touches and how much control you want around it. After a short intake you get one fixed price up front, not an open-ended bill. After that it can run on managed hosting from €39 a month. Model usage follows your volume and is listed separately in the quote. Which factors count: what an AI solution costs.

How do you start with an AI agent?

You start an AI agent on one real queue, with your own data. First we look at how the work runs today and which questions keep coming back. Then we build a working prototype on that one queue, so you can see whether it holds up before you commit to more. How long it takes depends on how many kinds of questions it handles and which systems it touches; we discuss your timeline in the intake.

Does this fit how you work?

Tell us where the work gets stuck. You hear from us within one working day, from the person who builds it.

Start a project