Amber Agents
Your data never leaves the EU

AI agents that
do the work
built for Europe.

Amber Agents answer your phone 24/7, resolve support requests, process orders and invoices, and answer your team's questions from your own documents.

We scope, prove and deploy with you, one process at a time.

30 minutes with one of our AI expertsNo obligation
ISO & GDPR certified
100% EU hosted
AI foundation models since 2018
Amber Agents · live conversation
Installer service line · incoming callLive
CallerHi, our heating stopped working this morning. Can someone come by today?
Handled by Amber Agent0:52DutchBooked in the planning system
Trusted by 5,000+ organisations across Europe
PhilipsPumaT-MobileDisney+ZDFTF1TransaviaUNICEFGemeente AmsterdamSodexoArvato
4.5 out of 5 stars4.5/5 Excellent, 1,000+ reviews G2: Users Love Us ISO 27001 · ISO 9001
10 days
from kick-off to a working agent on your real cases
the volume with the same team, who now own the exceptions
0
bytes of your data outside the EU. Sovereign hosting on STACKIT
Use cases

Agents for the work that eats your week.

These are the agents our clients usually start with. Every one of them reads your systems, acts within limits you set, and hands over to a person with full context when it should.

Voice

Customer support voice agent

Answers every call within two rings, 24/7, in Dutch, German, French and English. Solves order status, bookings, opening hours and FAQs from your systems. Hands the rest to a colleague with a summary.

Chat · e-mail

Chat & e-mail support agent

Same agent, same knowledge, on your website and in your inbox. Returns, invoices and delivery questions handled within the policy limits you set. Escalations become tickets with the conversation attached.

VoiceChat · e-mail

Appointment & booking agent

Books, reschedules and reminds by phone and chat, also at night and on weekends, straight into your calendar or practice system. For practices, garages, retail chains and service desks.

VoicePublic sector

Citizen service agent

Opening hours, waste collection, permits, passport appointments and status questions for residents, 24/7 in three languages. Answers come from published decisions and policy documents, with the source.

Back office

Order & invoice agent

Supplier orders, order confirmations, invoice matching and polite dunning processed end to end in your ERP and mailbox. Exceptions are routed to Finance with a proposal, never booked silently.

Internal

Company brain (Ask Amber)

One assistant over your documents, meeting notes, tickets and CRM. Every answer cites its source. Available in Slack, Teams, e-mail and by phone, and it hands tasks to your other agents.

Sales

Lead & meeting agent

Qualifies inbound leads and books the meeting within five minutes. After the meeting it drafts the recap and the proposal from your Amber Notes transcript and keeps the CRM up to date.

Operations

Operations & intake agent

Order completeness checks with OCR, spec validation, deadline monitoring and proactive customer updates for operations teams. Quality scoring of every delivered job, exceptions to the right person.

Your process not on the list? Every process gets scored in the scoping sprint, and the matrix decides where we start. See how the scoping sprint works →
Results

What clients see in the first quarter.

These are the numbers from our deployments in business and the public sector.

Municipality, 90,000 residents
60%
of calls fully handled

Citizen service line for opening hours, waste, permits and appointments, 24/7 in three languages.

Housing corporation
−40%
missed repair calls

Repair-request agent logs the issue, checks the tenant file and books the contractor slot, also at night.

University, 30,000 students
252 min
wait time in the enrolment peak

Student service agent in chat and phone answers 70% of enrolment questions; staff take the rest with context.

Consulting firm, 300 staff
2.5 hrs
saved per employee per week

Company brain over all documents, meetings and projects; 90% of employees use it every week.

Retail chain, 40 stores
+14%
revenue

Phone agent takes appointment bookings outside opening hours; 68% of booking calls handled without staff.

E-commerce brand
−76%
support tickets for staff

Order status, delivery and returns handled by voice and chat; an answer at 2 am costs the same as at 2 pm.

Wholesale distributor
0
extra finance FTE

Supplier orders, confirmations and invoices processed by an agent; 90% less manual purchase-order work.

Building materials supplier
92%
of incoming orders booked without a person

Order completeness check with OCR: e-mailed and scanned orders read, matched to the price list, missing details queried with the customer, then booked in the ERP.

Not sure which process to start with?

Thirty minutes with one of our AI experts is usually enough to find the first one.

How it works

Three steps to your first agent, with us at your side.

You don't need an AI team to start. We help you find the right first process, prove it on your own cases, and roll it out with your people.

Step 1 · Workshop2 to 3 weeks

Find the right first process

Together with your team we map your processes and score each one on effort and benefit. You leave with a clear shortlist and a named owner for every candidate.

Step 2 · Proof of concept10 working days

Prove it on your own cases

We build the first agent next to the people who do the work today, on 50+ real cases, and measure it against your baseline. Go or no-go on day 10, then go-live.

Step 3 · Rolloutquarter by quarter

Grow at your own pace

Once the first agent beats its baseline, we take the next process from the list. Your champions learn to run and improve the agents themselves; we stay one call away.

Where the return is high and the effort is low, we start.

The matrix puts every process on two axes: the benefit of automating it and the ease of doing so. The top-right quadrant is where the first agent goes. The rest is planned, or left alone on purpose.

Plan for later
Automate first
Leave as is
Quick wins
Ease of implementation → Benefit → Phone bookings Order status Invoice matching Complaints Lead follow-up Contract review
Days 1–2Deep diveShadow the real work, collect 50+ real cases, set the baseline and the approval points.
Days 3–5Build v1Configure the agent, connect your data and tools, define hand-over rules.
Days 6–8Test with realityThe agent runs on real cases next to your team. Evals on accuracy, tone and exceptions.
Days 9–10Go / no-goMeasured against the baseline. Go live, refine, or stop. Your call, with the numbers on the table.
Forward deployed

We don't consult. We think along, prioritise with you, and deploy.

Our forward deployed strategists and engineers sit with your team until the agent works, then hand over the keys. That is how you become AI native instead of dependent on us.

FDS

Forward deployed strategist

Owns the problem

Runs the scoping sprint, scores the processes, and works with your process owner until the prototype handles real cases. Defines the evals and the success metric. Accountable for the return.

Rule: nothing goes to engineering before it works on real cases.
FDE

Forward deployed engineer

Makes it production

Connects your systems, hardens the agent, sets up security, observability and cost control, trains your champions, and stays on call for the first month. Accountable for reliability.

Rule: nothing goes live before it beats your baseline.
You

Your champions

Own the work

Not engineers, but the people who know the work and its exceptions. They write the agent's instructions in plain language, own its knowledge, and run the next process with us. We train your people.

Result: your team runs its agents. We are one call away.
Agents Platform

One platform under every agent.

Studio, skills and connectors, knowledge and memory, voice, guardrails, evals, hand-over and insights. Best model per task, no vendor lock-in, and everything on sovereign European infrastructure.

Plain-language instructions, tested before they shipPlaybooks, a skills library and a proving ground with simulated callers and regression tests.
Guardrails and a supervisor on every answerPolicy limits, approval points for irreversible actions, PII masking, a full audit trail.
Sovereign by designSTACKIT sovereign cloud, EU model endpoints, ISO 27001. Swap models without moving your data.
console.amberagents.eu / vandijk-groothandel / agentsAll systems live

Agents

New agent
1,842invoices this month
96%+2booked without a person
9 sper invoice, was 12 min
€0.31cost per invoice
AgentStatusTodaySolvedRelease
Invoice agentDigital · ERP + mailboxLive8796%v9
Order intake agentDigital · OCR + ERPLive6492%v6
Support lineVoice · NL, DE, ENLive21271%v14
Dunning agentDigital · shadow modeShadow0v2
Invoices processed · last 7 days
Monitors
Matching accuracy98.4%
Exceptions to Finance4.1%
Duplicate invoices caught3
Open alerts0
Why Amberscript

Eight years of AI in production for 5,000+ European organisations.

We have built our own proprietary foundation AI models since 2018. Evals, review loops and exception handling are exactly the discipline agents need. Now we bring it to your processes.

2018
own AI engines, built in Amsterdam
5,000+
business customers
2M+
user accounts
99%+
quality with AI plus humans in the loop
100+
people in Amsterdam and Berlin
Security

Built and hosted in Europe.

Your agents handle sensitive data. That should stay yours.

EU data centres only

Production in Frankfurt today. Amber Agents run on STACKIT, the sovereign cloud of the Schwarz Group, for public sector and regulated clients. Single-tenant on request.

Models hosted in the EU

Language and speech models run on EU endpoints under our contracts. Your data is never used to train models and never leaves the EU. Zero retention with every provider.

Your data stays yours

Data processing agreement before day one, encryption in transit and at rest, deletion on request or automatically after six months, and a full audit trail of every action an agent takes.

Certified operations

ISO 27001 and ISO 9001 certified, GDPR and NIS2 compliant processes, Trusted Partner Network member. Logging, human oversight and transparency built in for the EU AI Act.

No foreign access to your data

US cloud and AI providers fall under the US CLOUD Act. European hosting under European contracts keeps your citizen, customer and employee data out of that reach.

Procurement and compliance ready

GDPR, NIS2 and the EU AI Act ask where data lives and who can reach it. A fully European stack answers those questions before they are asked. DPIA support and sub-processor list on request.

Pricing

Pay per agent per month. No surprises.

Every agent has a fixed monthly price by type and size. Model usage, hosting, observability and updates are included.

Digital agent

Tickets, orders, invoices, intake, chat, e-mail. Works in your systems and mailboxes.

S500 runs a month€395/mo
M2,000 runs a month
L8,000 runs a month
  • A run is one completed job: one ticket resolved, one order processed, one invoice matched
  • Standard connectors, approval points and audit trail included
  • Unlimited channels: chat, e-mail, Slack, Teams, WhatsApp

Company brain

Ask Amber over your documents, meetings, tickets and CRM. Unlimited users, in Slack, Teams, e-mail and by phone.

Unlimited users, fair use€295/mo
+Included with any live agent
  • Connect Google Drive, SharePoint, OneDrive, Amber Notes, your ticketing and CRM
  • Every answer cites its source; knowledge gaps flagged weekly
  • Hands tasks to your digital and voice agents
FAQ

Frequently asked questions.

Where does our data live, and who can reach it?
In EU data centres only: Frankfurt today, and STACKIT, the sovereign cloud of the Schwarz Group, for Amber Agents. Models run on EU endpoints under our contracts, with zero retention. Nobody outside the EU can be compelled to access it. You get a DPA, a sub-processor list and a full audit trail.
Which AI models do you use?
The best model per task: Claude, Mistral and open-weight models for language, and the best speech-to-text per language, including our own engines for Dutch and German. There is no single-vendor lock-in; models can be swapped without moving your data.
How long until an agent is live?
Ten working days to a working agent on your real cases, with a go or no-go decision against your baseline on day 10. Weeks three to six are for hardening, security review, champion training and go-live. The scoping sprint before that takes two to three weeks.
What happens when the agent doesn't know?
It says so and hands over. Every agent has hand-over rules: complex or emotional cases go to a person with the full conversation and a summary, and anything irreversible waits for approval. Escalations become tickets, and the gaps are flagged so your champion can close them.
Do you replace our people?
The agents take the routine; your team owns the exceptions. In our own support team, the same headcount now handles three times the volume. Choices on redeployment and reskilling are made with you and agreed up front, in the scoping sprint.
Can it work with our phone system, CRM and ERP?
Yes. Standard connectors cover common telephony, ticketing, CRM, ERP, calendar and document systems. Anything else, your forward deployed engineer builds once as a project, after which it becomes a standard connector.
What do you need from us?
A process owner who can spend a few hours a week during the proof of concept, access to real cases and the systems involved, and one champion per department who knows the work. No AI team, no data scientists.
We are a municipality. How does procurement work?
A packaged citizen service agent is available through our municipal partners, priced per municipality and typically under direct-award thresholds. ISO 27001, GDPR, NIS2 and sovereign hosting on STACKIT are standard. DPIA support is included.
Next step

Which process would you automate first?

Book 30 minutes with one of our AI experts. They will ask about your processes, tell you honestly whether an agent fits, and show what the first ten days would look like. If it doesn't fit, they will say so.

30 minutes with an AI expert
Amberscript · Amsterdam and Berlin

What you will leave with:

  • A first read on which of your processes fit an agent
  • A realistic estimate of effort, benefit and price
  • A proposed plan for the scoping sprint, if it makes sense