AI that works inside your own system.
Not in a tab next to it.
For small and mid-sized companies that want more than ChatGPT in a separate window. I build AI into the software your team already works in: an assistant that searches your own orders, quotes and documents, prepares drafts and changes nothing without approval. One scoped problem at a time, with a fixed price after one call.
Why AI stalls in most companies.
ChatGPT in a tab next to it
Staff copy client data into a chat window and paste the answer back. It works, but nobody knows which data ends up where.
The pilot that never finished
There was a demo that impressed everyone, and then nothing happened. It was not connected to your systems, so nobody could do their work with it.
Answers based on nothing
A language model that is allowed to guess names a price, a stock level or a client that never existed. In business software that is a wrong quote.
Looking up the same things daily
What is the status of that order, what did we agree with this client, which quotes are still open. The answers are in your system, but someone has to click them together every time.
Retyping from email and PDF
Requests, receipts and purchase orders arrive as text and go into the system by hand. That is where the hours and the typos are.
No idea what it will cost
Usage is billed per question. Without a limit and a meter, one enthusiastic user is an open-ended bill.
From idea to something your team uses.
Pick one problem
Not 'something with AI', but the work that costs hours every week and whose result someone can check. That conversation decides whether it becomes useful.
Connect it to your own data
For every kind of question a function that fetches the real data from your database, with the permissions of the user asking. No fetched data, no answer.
Proposing instead of executing
If the assistant wants to create or change something, that appears on screen as a proposal. Only when someone approves does anything happen.
Measure and cap
Usage per organisation, a daily limit and an on-off switch for the whole thing. You see what it costs before the bill arrives.
First check whether an existing tool will do
ChatGPT and Copilot are good at writing, summarising and thinking along, and they cost a fraction of custom work. If it is about separate tasks of individual staff, a subscription with clear rules on what may go in is the sensible route and I will say so.
Custom starts to pay off once the AI needs your own business data, or once the answer has to lead to something.
It needs your own data
Orders, quotes, projects and stock live in your system, not in a language model's head. An assistant that cannot reach them is guessing.
Something has to happen
Creating a task, preparing a quote, forwarding a ticket. From a separate chat window that stays copy and paste.
Not everyone may see everything
An engineer sees their own jobs, management sees the total. The assistant gets the same permissions as the person asking.
Your clients ask the question
Then the assistant belongs in a [client portal](/klantportaal-laten-maken), where each client only gets their own data back.
Where AI saves time in a company
The applications that stick are rarely spectacular. It is work done by hand today, that comes back often and whose result a person can judge in a few seconds.
Calculation is not one of them: a price or a margin goes through a calculation function, not through the model. If your data is still spread over separate packages, the integrations come first.
Searching your own data
A question in plain language instead of clicking through five screens: which quotes are open, what was delivered to this client last month.
Preparing drafts
A reply to a client question, a quote text or a summary of a file, filled with the real data. Someone reads it and sends it.
Reading documents
A request, receipt or purchase order from email or PDF becomes a filled-in form. The user checks and saves.
Sorting and routing
Incoming messages get a subject and an urgency and land with the right person.
How it goes now, how it goes with AI
The same work, before and after. No job disappears: what goes is the looking up, retyping and preparing, and what stays is someone who checks and decides.
| Work | How it goes now | With AI in your system |
|---|---|---|
| Looking up an order status | Clicking through five screens and calling the client back | One question in plain language, answered with the real data |
| A request from email or PDF | Retyped into the system by hand | A filled-in form that someone checks and saves |
| Quote text or client reply | Written from scratch or copied from an old document | A draft with the right client, products and agreements, ready to read through |
| Incoming messages | One inbox where someone forwards everything | Subject and urgency recognised, straight to the right person |
| Cost of AI usage | Separate subscriptions with no view of what goes in | Usage metered per organisation, with a daily limit |
What happens to your data
Only the question and the fetched data needed for that one answer go to the provider you choose: Anthropic, Google or OpenAI, or a European party if your sector requires it. How the assistant itself is put together is on AI assistant in your own software.
- Keys stay on the server, never in the browser
- Conversation content is not stored by default
- Permissions enforced per role and per organisation
- Claude, Gemini or GPT, switchable through a setting
- Daily limit per organisation
- One switch to turn the whole thing off
Step plan: integrating AI in five steps
The first step decides the most. An application that solves one scoped problem is usable in weeks; 'something with AI' is still a pilot after six months. How a project runs beyond that is on how I work.
| Step | Duration | What you get |
|---|---|---|
| 1. Call | Half an hour | One scoped problem, and an honest answer on whether AI is the right tool for it |
| 2. Proposal | After the call | Fixed price, scope and planning in writing |
| 3. Build | 3 to 6 weeks | An assistant inside your existing application, connected to your own data, that your team tests before it goes live |
| 4. Live | At the end of the build | Permissions per role, daily limit and cost metering set up; repository and hosting in your name |
| 5. After that | Ongoing | The next application once the first one works, and maintenance for a fixed amount per year |
What it costs and how long it takes
The build is a fixed amount you hear up front. What that amount is depends mostly on where your data lives; these are the three forms that come up most, with amounts excluding VAT. Usage you pay directly to your provider, capped with a daily limit.
How pricing works is on pricing. If there is no application yet for the AI to live in, it starts at custom software development. On demos there is a working assistant to try before you decide anything.
Assistant in an existing application · € 2,950 to € 5,500
Your system is already there and the data sits in one database. The assistant gets a function per kind of question and the permissions of the user. Lead time three to six weeks.
AI on data from several systems · € 4,500 to € 8,500
The answers are spread over your ERP, bookkeeping and webshop. Then the integrations are built first and the assistant goes on top. Lead time five to nine weeks.
New application with AI built in · € 6,500 to € 12,000
There is no system yet for the AI to live in. It becomes a software project in which the assistant is one of the parts. Lead time seven to twelve weeks.
An assistant inside your own application
Pick a question the way an employee would ask it:
Every answer comes from a tool that fetches your own data, with the permissions of the person asking. Creating and changing never happens by itself.
AI in practice in the real world
Clients rate me 5 stars on average.
“I am extremely happy with the result of my new website for DoubleStreet Australian Cobberdogs, made by Webframer. From first contact, Webframer radiated expertise and creativity.”
“Webframer helped me enormously with the Veldvisuals website. It is not only good to spar with Bas about building the site, but also in the aftercare and whenever changes have to be made. Recommended!”
“Great to work with. Listens carefully to what we want. When we ask for changes, they are made expertly.”
“Really pleasant to work with and short lines of communication!”
“The website looks great!! Definitely recommended!”
“Excellent! Fast service and good contact!”
Translated from the Dutch originals on Google.
I am Bas Voets. Since 2021 I have been building websites, client portals and business software for small and mid-sized companies, from Hazerswoude-Dorp in the Netherlands. From first call to launch you talk to me, not to an account manager.
How do you start integrating AI into your business?
Does my business data go to an AI company?
Can the AI make things up?
Can the AI change things in our system on its own?
Is ChatGPT or Copilot not enough?
What does AI integration cost?
How long until it is live?
Is my data used to train AI models?
Where is my data stored?
What does the EU AI Act mean for my company?
Does AI replace my staff?
Tell me what you want to build.
A half-hour call. You get an honest answer about what it costs, how long it takes and whether building it is smart at all.