Questions
Short, honest answers about how WinSphere picks up the work nobody gets to at firms and agencies with 10 to 100 people: what it costs, how it works, and why not do it yourself or through another agency. No sales talk, just the questions we hear most.
WinSphere, founded by Rafael Matabadal in The Hague, is the AI and automation agency for owners of firms and agencies with 10 to 100 people: accounting firms, recruitment agencies and B2B service providers. We don't touch their work, we pick up what nobody gets to: first we run their numbers, then we build, then we keep it running.
What you get
A plan with the numbers run on your own figures, even if that is where it ends.
Workflows that run in your own systems, not a black box.
Every four weeks, in black and white, what it saves.
The final judgment stays yours: we automate the work, not the judgment.
In this overview
9 questions
An AI and automation agency builds the software that takes over recurring work inside the systems a company already uses. We do that for firms and agencies with 10 to 100 people, and we first work out whether it pays off.
Workflow automation is a system that carries out a fixed sequence of steps by itself, from the moment something comes in to the moment it is handled, without anyone in between. In our case that means, for example: requesting documents, sorting what comes in, updating the file, and only showing you what is ready for review.
The work that comes back every week, that nobody is assigned and that appears on no invoice: chasing documents, retyping intake, follow-up, the monthly report. It costs hours and brings in no revenue, so it shows up in no overview.
In three steps. First thirty minutes in which you tell us how a normal week runs and we ask the follow-up questions; no presentation, no proposal. Within a week you get a written plan from us: which processes swallow the most hours, which can be automated and which cannot, and what that saves per year in euros, calculated with your own figures. We go through that plan together in a second call, and it is yours, even if that is where it ends. Only after that do we build, in four-week cycles, with a baseline up front so that every four weeks you see in the numbers what has changed.
You can, and sometimes that is exactly right: for a simple connection a tool like Zapier or Make is often enough. Where it goes wrong is with workflows that have to keep working as you grow: nobody watches over them, error handling is missing, and the handiest person in the office becomes the only one who knows how it fits together. We build the way a developer builds: from a brief and a knowledge base, with code where that holds up better. n8n is one of our tools, and we are just as happy to build on your own n8n. That is usually our advice too, because then you own it, with logging and error handling underneath, and maintenance does not hang on one person.
That can work too, and increasingly well. Where it goes wrong: a system that picks an API without error handling or security, and that nobody understands any more once the person who built it leaves. We use the same tools where they are faster, but we test every build on maintainability, security and measurement before it goes into production.
Yes, and that is the first step we take. AI or automation on top of messy data mostly automates the chaos, not the work. We first put your data in order and connect the systems, and only then automate the rest. That way you avoid ending up with a system that does the wrong things faster.
No. We don't touch the way you work: the professional judgment, the client contact and the signature stay with your team. We build the automation around your existing process, in your own systems, so you don't have to learn a new package before the work gets picked up.
That differs per firm, so we do not name an end date up front. The order is fixed, though. First, thirty minutes in which you tell us how a normal week runs. Within a week after that you have a written plan: which processes eat the most hours, which ones can be automated and which cannot, and what that is worth per year. We go through that plan together, and only then do you know how much there is to do and in what order. After that we build in four-week cycles, with a baseline up front. Every cycle delivers something that runs, so you are not waiting for one big handover.
3 questions
The thirty minutes and the plan cost you thirty minutes and your own numbers. What an engagement costs after that, you hear in the second call, before you decide anything: set against what the plan works out, per four-week cycle, and never as an hourly rate. We also say so if the math does not add up for you. A rule of thumb up front: if the calculation comes out below a few hundred hours a year, an engagement will not pay for itself at your firm, and the plan will say so.
We work best with firms and agencies of 10 to 100 people: that is where the most time leaks into work nobody gets to or work your team is still doing by hand, and where the gain is biggest. You don't have to be big or technical. We start with the pile that is most in the way right now and build out from there.
An extra employee keeps doing the same manual work, at a fixed cost and without scaling with the workload. We take the manual work away: your team does more without growing, and every four weeks you see in the numbers what that delivers.
2 questions
A generic AI agency often sells an AI strategy or a demo that stands apart from your systems. We know the vocabulary of your sector, first work out what the work nobody gets to is costing you, and only build once that math adds up. Not a strategy document, a working system in production.
Fair question, and the answer is not a promise but the way we build. There are capable people in our network who can take the work over straight away. The systems currently run on our own n8n, but we build them on yours just as readily, and that is what we recommend: then it sits with you and it is yours. And everything we build is documented: how it works, in documentation that both a person and a language model can read. At handover you get that documentation, and where it helps a video walkthrough with it.
Your data stays in your own systems; we connect only what is needed and set up processing in line with the GDPR. If we use AI models in a workflow, we also weigh what the AI Act requires for that. More on how we handle this is on the AI Act page.
WinSphereAI and automation for firms and agencies with 10 to 100 people. First costed out, then built, then kept running.