NinA AI Agency

AI Agents · digital colleagues

AI agents that do real work

An AI agent reads the email, looks things up in your ERP, drafts and proposes. We build agents that work in your own systems, know when to ask, and leave the final click to a human.

  • 192 hours a month back at Van Berkel
  • Human in the loop
  • Data in the EU

Purchasing agent

live
An order confirmation from Bakker BV just came in. Does it match our order?

reads attachment confirmation_8841.pdf

looks up purchase order PO-2210 in Exact

Two differences. Line 3 is priced at € 14.20 instead of € 12.80, and delivery is a week later: 14 November.
Email them about the price. The date is fine.

creates draft email in Outlook

Draft is ready in your outbox. Want to read it before I send?

Trusted by

  • Schiphol Group
  • ASML
  • Politie
  • Action
  • Boskalis
  • Gemeente Lelystad
  • KplusV
  • de nieuwe arts
  • Last Mile Solutions
  • Nsecure
  • DA Drogist
  • De Baak
  • EasyToys
  • KernBouw
  • De Koffie Jongens
  • Wens Chalets
  • Cascando
  • Qcore
  • Avictus
  • MoveBuddy
  • Nomilk2day
  • Red Panda Works
  • Cakeprint
  • TZA

01 / What is an AI agent?

Not a chatbot. A colleague that takes the steps itself.

A chatbot answers a question. An AI agent gets a task and works out how to finish it: what information it needs, which system it's in, what the next step is. It uses tools like your ERP, mailbox or knowledge base, and reports back.

The difference with plain automation: a workflow follows fixed rules, an agent can handle variation. An order confirmation that looks slightly different, a customer asking three things in one email. A fixed rule gets stuck there; an agent understands what's meant and asks when in doubt.

  • 01Plans its own steps

    It gets a goal, not a script.

  • 02Uses your tools

    ERP, CRM, email, SharePoint, planning. Only where it's allowed.

  • 03Remembers context

    Earlier orders, agreements with a customer, your house rules.

  • 04Knows when to ask

    When in doubt or above a threshold, it asks a human.

02 / Where it pays back

Three places where an agent pays for itself fastest.

Every business thinks it's unique. The patterns are recognizable. See 20 examples by department.

  • Expert knowledge

    Documents and reports

    Calculations, advice and compliance documents. The agent drafts from hundreds of examples, the expert approves.

  • Retyping

    Requests, orders and invoices into your ERP

    Email, PDF and portals read automatically and placed in your system.

  • Volume

    Customer service

    Agents that prepare the answer, route tickets and reply in multiple languages. Humans handle exceptions.

03 / The AI floor plan

Where does an agent fit in your company?

Click through the office. For each department you see what AI takes over and which role people keep. The numbers come from real projects.

Click a department

What AI takes over here

Customer service

  • Draft a reply with the order detailsHuman approves
  • Label and route ticketsFully AI
  • Call the difficult customerAI assists

In practice

30% → 80%

of customer questions get a prepared reply (enterprise client)

More on AI in customer service →

04 / How we build an agent

From idea to colleague in six steps.

  1. 01

    Choose

    Quick Scan

    Which process, what volume, what it delivers. We only build when the numbers add up.

  2. 02

    Design

    Week 1

    What the agent may do itself, what it proposes, which systems it touches.

  3. 03

    Build and connect

    Weeks 2 to 4

    In n8n and custom code where needed, in your environment.

  4. 04

    Test on real work

    Weeks 4 to 6

    Hundreds of real cases from last month, measured before anyone relies on it.

  5. 05

    Live, with a human

    Then

    At first a human approves everything. What goes well gets more room.

  6. 06

    Measure and adjust

    Every month

    Hours saved in euros, errors, exceptions, and what's next.

05 / Safe

An agent with access to your systems needs agreements upfront.

  • A human is always in the loop

    No agent that communicates or books without control.

  • Data stays in the EU

    Your infrastructure or our servers with a European provider. Working to ISO 27001 and NEN 7510.

  • No lock-in

    Code, prompts, data and documentation are yours. The model is swappable.

07 / Questions

About having AI agents built.

What does it cost to have an AI agent built?

We build agents within the AI Partnership: from € 4,500 per month for 30 hours of our team, at one rate of € 150 per hour. What it costs and returns is calculated in the AI Quick Scan. More in What does an AI agent cost?

How long until an agent is live?

Usually four to eight weeks from design to a first production version, with a human approving everything.

Does an agent replace our employees?

No. An agent takes over the retyping, searching and summarizing. People do the judgement, the contact and the exceptions. The goal is to grow without hiring, not to work with fewer people.

Which colleague do you hire first?

  1. 01

    Intro call

    Fifteen minutes, no strings. Where does work get stuck and what do you want to achieve?

  2. 02

    AI Quick Scan

    One day on site. Within a week, a business case and a roadmap.

  3. 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