AI agents that carry out tasks in your own systems.

An agent carries out the task itself in your own systems, for example processing email or preparing bookings. You decide in advance what it may do.

  • Works in your own systems
  • Asks for approval where you want it
  • Every step in a log

The difference between a chatbot and an agent.

A chatbot tells you what to do. An agent does it, with access to the systems you already use.

ChatbotGives an answer
Create an invoice for Bakker Installatie, for last week's hours.

To create an invoice in your accounting package you can follow these steps:

  1. Go to Sales and choose New invoice.
  2. Look up the customer.
  3. Add the hours worked as lines.
  4. Check the VAT and save.

Let me know if you have any more questions.

ResultYou still have to do it yourself.
AI agentCarries out the task
Create an invoice for Bakker Installatie, for last week's hours.
Calendar and work orders3 work orders found, 11.5 hours
CRMCustomer and rate retrieved
AccountingDraft invoice F-0418 created
CheckHours and rate match the contract

Invoice F-0418 is ready: 11.5 hours at €150, total €1,725.00 excl. VAT. Shall I send it?

ResultThe invoice is ready. You click send.

Work an agent takes over from you.

Agents are good at work where you have to read, look up and fill in. Tasks that vary too much for ordinary automation, but keep coming back.

Inbox

Sorting and answering email

Reads incoming email, files it with the right customer and writes a draft reply based on your earlier replies.

OutlookCRM
Administration

Processing invoices and bookings

Takes purchase invoices from email, reads the lines and prepares the booking in your accounting package.

EmailAccounting
Sales

Preparing quotes

Turns a request into a quote with the right products, prices and terms. You check it.

CRMPrice list
Customer service

Questions about orders

Answers questions such as "where is my order" with the real status from the system.

WebshopCarrier
Check

Checking documents

Compares contracts, work orders or purchase orders and marks what differs.

DocumentsERP
Planning

Scheduling appointments

Finds a time that suits customer and engineer, and puts it in both calendars.

CalendarPlanning

You record what an agent may do by itself.

An agent works for you, and you set the rules. In advance we record together what it may do itself and what it may not.

You set the limits

Per action we record what the agent may do itself, where it asks for approval and what is never allowed.

Everything in a log

Every step the agent takes is recorded. You see what it did and why.

Your data stays yours

We choose models and hosting that fit your requirements. Your data is not used to train models.

What may the agent do?Does it itselfAsks for approvalNever
Reading email and documents
Writing a draft reply
Looking up data in the CRM
Sending email to a customer
Creating a booking or invoice
Making a payment
Deleting data

An agent gets more room step by step.

You do not give a new colleague all the keys on day one either. An agent gets more room as it shows that it can do the work.

Low autonomyChoosing one task

We start with one task that takes a lot of time and can be clearly scoped.

Everything past a personWatch

The agent does the work, but everything goes past a colleague. That way you see how good it is.

Growing trustLetting go where possible

What keeps going well, the agent may do itself. The rest keeps going past a person.

Independent where possibleExpand

If it works, we add tasks. Step by step, with the same approach.

Frequently asked questions about AI agents.

Is your question not listed? Call +31 10 318 0515.

What is the difference with ChatGPT?

ChatGPT answers a question. An agent carries out a task in your own systems, such as your email, CRM and accounting, within the limits you set.

Which AI models do you use?

We choose that per task. Sometimes a large language model through a business provider, sometimes a smaller model on our own servers. The most important question is where your data is allowed to go.

What if the agent makes a mistake?

That is why every agent starts supervised: a colleague approves everything. Only what keeps going well may it do itself. And every step is in the log.

What does an AI agent cost?

For the build you get a budget in advance. In addition you pay the costs of the AI model per use. We make those clear in advance.

Does it work with our systems?

If a system has an integration or API, the agent can work with it. That applies to most modern packages. If it does not, we look for a solution together.

The coffee is already on.

Bring your question, your process or an idea that is not finished yet. In a first conversation we map out your situation and you get honest advice on the next step.

Freshly made

How do you take it?