AI specialist for Dutch SMB
We build and deploy AI where it provably removes manual work.
We work with leadership and IT teams of mid-sized organisations who take AI seriously, not as a gimmick.
Who we are
Vectel is a broad IT partner for Dutch SMBs. AI automation is our most substantial practice: we pick the right tool for the problem and are honest when AI is not the answer.
Our AI practice
Five ways we deploy AI, all with a measurable outcome.
Microsoft Copilot implementation and adoption
Roll out Copilot, set up governance, and make sure teams actually use it.
Book a Copilot conversation →AI automation
Speed up repetitive processes, unstructured data handling, or decision-making with LLMs and agents.
View the automation practice →AI integration in existing systems
LLM capabilities inside your CRM, ERP, or customer portal, with attention to latency, cost, and governance.
Discuss the integration →AI audit and readiness
An objective view of where AI delivers value in your organisation and what needs to be sorted first.
Start the scan →AI training for teams
Practical sessions where teams learn to use AI responsibly in daily work.
Plan a training →
Our approach
Five phases, no consultancy jargon.
- Step 1Discovery (1 to 2 weeks)
What problem are we solving, for whom, and how do we measure it?
- Step 2Scope (1 week)
What is out of scope, what data do we need, what risks do we accept?
- Step 3Pilot (4 to 8 weeks)
Small, measurable, with time to adjust before we scale.
- Step 4Measure (ongoing)
What changes, for whom, how much manual work actually disappears?
- Step 5Scale
Only when the pilot delivers evidence, not before.
Are you ready?
Five criteria we assess before starting a project.
Data quality
Is the data findable, clean, and do you know who owns it?
Governance
Who can do what, what is the GDPR legal basis, and in what context is it deployed?
Leadership
Is this supported by the board or management team, or is IT on its own?
Use case clarity
Concrete problem or 'something with AI'? The second is a red flag.
Integration context
Which systems need to connect, and on what timeline?
Common mistakes
AI on messy data
We typically see that data needs to be sorted first, otherwise AI performs at best the same as the old process.
Choosing tooling before the use case
The tool choice should follow from the use case, not the other way around.
Pilot without success criteria
Without agreed-upon metrics up front, you will not know afterwards whether it worked.
Adoption as an afterthought
A great AI feature that nobody uses is not a win.
Compliance at the end
Bringing in GDPR and NIS2 only at the last moment costs more than addressing them upfront.
Tooling choices
We choose tooling after discovery, never before.
- Microsoft Copilot
- When the organisation already runs M365 and the use case fits within Office.
- Claude or GPT via API
- When custom work or integration outside M365 is needed.
- Custom agent framework
- For multi-step processes with domain logic and autonomous steps.
- No AI
- When a rule engine, form adjustment, or better search is the real answer.
Data quality and governance
AI only delivers value when data is in order and governance is clear. For organisations under NIS2, that starts with a basic inventory.
Take the NIS2 scan →Cases from our practice
Anonymous examples, because we do not discuss clients without permission.
Document extraction
Invoices, receipts, contracts and email attachments parsed automatically.
Purchase invoices and receipts retyped into the accounting package eat hours a month and leave typo errors that have to be tracked down later. An AI extractor reads the structured fields, checks them against the order line, and posts them through when the match holds.
Read more →Mail triage for the shared inbox
info@ and support@ sorted, summarised, routed. A human decides what goes out.
The shared inbox is the quiet pain of many offices. Three people read the same mail, nobody picks it up because they think someone else will, and the customer waits a day for something that could have been done in five minutes.
Read more →Knowledge portal over your own documents
One search box over SharePoint, handbooks and wiki. Answer with a source link or an honest "no idea".
New joiners ask the same questions for three months that are already answered in a document nobody can find. An internal search that knows which page it does not know is worth more than a chatbot that guesses.
Read more →Quote flow for sales
From a customer request to a draft quote in minutes, not days. A human hits send.
A salesperson gets three requests in on Monday. By Thursday they are still sitting there, because copying from earlier quotes, looking up price lists, and polishing the branding is the work that gets pushed aside between customer calls.
Read more →Meeting summaries that actually exist
The kind of meeting whose minutes never get written, now gets a finish line.
In most SMB meetings nobody writes minutes. Not from ill will, but because the person who would type them is part of the conversation. A short summary afterwards is worth more than a forty-minute transcript nobody reads.
Read more →Lead classification with reasoning
A readable priority list per lead, with arguments, not a black-box score.
A lead form on the site, a whitepaper download, a conference badge scan. At the end of the week eighty names sit in the CRM. Which five should sales call today, and why?
Read more →
Outcomes are indicative and project-specific.
Where we operate
We work from Veenendaal for clients across the whole of central Netherlands.
Frequently asked questions
What does an AI project cost?
Depends on scope and data state. After discovery we give a substantiated range, not a fixed price upfront.
How long does an AI implementation take?
A typical pilot runs 4 to 8 weeks. Scaling after that depends on scope.
Does AI work for small businesses too?
Yes, provided the use case is sharp. Being small is no blocker, vagueness is.
What if our data is not clean?
Then the project starts with data, not with AI. Otherwise you get an expensive parrot.
Which tools do you use?
Depends on the use case. We work with Microsoft Copilot, Claude, GPT, and custom agent frameworks where appropriate.
What GDPR risks does AI introduce?
Mainly around purpose limitation, legal basis, and transfers. We address this in the scope phase, not at the end.
Do you work together with our current IT partner?
Yes, often. Our role is then AI specialist within your existing IT architecture.