What is running
At a search agency, a search agent we built is running. Our own exclusion list merges new replies from HeyReach, Instantly and HighLevel into one clean list every day at 05:00, so nobody is approached twice.
WinSphere itself runs on the systems we build for clients.
exclusion list
Live
Our own operation
Runs daily
What else is running
The five routes
Including the routes we do not appear in. So you first get when you do not need us, and only after that the comparison.
01
Leave it
When this is the best choice
The best choice when the pile is small, or when something bigger is going on: a merger, a software migration, someone leaving. Anything you build now will not fit six months from now, and waiting is not delay but sequence.
What it costs
The leak runs on for another year, and because the amount never appears on an invoice, nobody sees it.
02
Hire someone
When this is the best choice
The best choice when the work calls for judgment, client contact or professional expertise, or when the volume is so irregular that no two weeks are alike. That is where you want a person. It is also the route if you want to keep it fully in house. Just know that a new hire usually does not bring this knowledge along, and that afterwards it hangs on a single person.
What it costs
Fixed cost, every month, busy or not. The manual work stays manual, just faster. And if that person leaves, the process leaves with them.
03
Build it yourself in a tool: n8n, Zapier or Make
When this is the best choice
The best choice when it is one or two integrations between tools that both have a clean API, and someone at the office enjoys it and has time for it. Zapier is fastest for standalone connections, Make sits in between, n8n gives the most control and can run in your own environment. We work with the same tools, and where a tool falls short we write code.
What it costs
Not the building, but the upkeep. An API changes and the flow stops without any alert. You pay per task executed, so the bill grows with your volume. And the colleague who built it has other priorities three months later.
04
Build it yourself with AI: Claude Code or Cursor
When this is the best choice
The best choice when you are technical yourself or want to become technical, when it is an internal tool with no client data in it, and when it is allowed to break without anything going wrong at a client. Then it moves fast and you learn a lot. We build with the same tools ourselves.
What it costs
Four things you only see later: nobody has it as their job to repair it when an API changes, the tool picks a route that works and not necessarily the safest one, there is no baseline so nobody can show afterwards what it delivered, and it runs on one person’s account, laptop and memory.
05
Hire an agency
When this is the best choice
The best choice when you want to put down one big project in one go and there is someone internally who watches over it. An agency with plenty of hands builds a lot at once faster than we do.
What it costs
A project often starts from what can be built instead of from what it delivers, and after delivery it has become your system.
WinSphere
And us? We pick up the work nobody gets to, without your way of working changing: first costed out on your own figures, then built, then kept running.
Side by side
The same seven questions for every route. No amounts, because any amount we would name about your situation is a number without a source; the amount that does matter is one section further down.
Advantage of this route
Risk of this route
No marker: neutral, and that is intended
Leave it
Hire someone
DIY, in a tool
DIY, with AI
An agency
WinSphere
What it costs you up front in your team time
+
Nothing, and that stays that way
Recruiting, onboarding, coaching
Evenings of a colleague
Evenings, plus figuring it out
Workshops and alignment
+
Thirty minutes
How fast you know whether it adds up
-
Never; the leak is on no invoice
After a few months, on gut feel
Only once it runs
Only once it runs
Usually only after delivery
+
Within a week, on paper
Whether your way of working and your systems have to change
+
No
+
No
Sometimes; the tool has a say
Often; something new is added
Often another package on top
+
No; we pick up what nobody gets to
Who keeps it running when it breaks
Not applicable
That person, if they are around
The colleague who built it
The colleague who built it
Whatever the contract says
+
We do, with logging and error handling
Who watches security and the GDPR
Stays as it is now
+
Your own policy
The tool covers the basics, the setup is yours
-
Whatever the tool picks
Ask about it; it varies
+
Recorded who processes what
What happens if the person who built it leaves
Not applicable
-
The process leaves too
-
Nobody knows what is running anymore
-
Nobody can read the code
You have a contract
+
Documented, and in your own environment if you want it there
What you are left with if you stop
Nothing changed
A vacancy
Loose flows, no overview
Code nobody knows
Whatever the contract says
+
The written plan, and it is yours
Every route has a situation where it is the best choice; those are above. This table is about what happens after that. What happens at our end if something happens to us is in the questions at the bottom of this page.
Building it yourself
This is the question we get most often, and the honest answer is not "leave it to us". We build with n8n, with Claude Code and with Cursor ourselves, and we write plain code where that holds up better. The difference is not in the building. It is in what belongs around it the moment client data goes in and it has to work on Monday.
Most owners could figure this out themselves. They just do not: running the business comes first, and internal development keeps slipping. That is exactly why WinSphere exists.
Where it usually goes wrong
-
Maintenance. The flow works on the day it is built. If an API changes it stops, often without any alert, and it is nobody’s job to repair it.
-
Security and the GDPR. An AI tool picks a route that works, not necessarily one that is safe. Keys stay hard-coded and data passes through a service nobody has vetted.
-
Measurement. Without a baseline nobody can show afterwards whether it delivered anything, and it stays a feeling.
-
Continuity. It runs on one person’s account, laptop and memory.
-
Volume. With a tool that charges per task executed, the bill grows along with the growth you were after.
When building it yourself is fine
+
It is one or two integrations between tools that both have a clean API.
+
There is no client data in it, or it stays inside a system you already use.
+
It is allowed to break without anything going wrong at a client.
+
There is someone who enjoys it, has time for it, and still has that time next quarter.
+
You want to learn it yourself. That is a good reason, and then outsourcing would be a waste.
If you recognize this list and it still runs fine at your firm, you do not need us. If you recognize it and have never looked at it this way, that is exactly what comes out in the 30 minutes.
Do the math
Six routes come with one number, and that number is the same for all six: what does it cost that this work keeps sitting there? Most firms do not know, because those hours never appear on an invoice.
A worked example from accountancy.
€168,000
per year, at ten employees
€16,800 extra revenue per employee per year
Utilization in this example
Now
75 percent
In this example
85 percent
Benchmark: 73 percent on average, 84 percent top quartile.
At accounting firms the utilization rate averages 73 percent and sits at 84 percent in the top quartile, at an average hourly rate of 105 euros (Simplicate Accountancy Benchmark, third quarter of 2025, almost 400 firms of 5 to 120 FTE). An employee who goes from 75 to 85 percent billable delivers roughly 16,800 euros of extra revenue per year at 1,600 workable hours.
This is an example with assumptions, from a sector that is not necessarily yours. It only works if there is work waiting to fill those hours, and not every firm gains ten percentage points. That is why in the plan we work it out with your own rate, hours and staffing, and we also say so when it does not add up for you.
Waiting another year costs you that leak one more time, and because the amount is written down nowhere, nobody notices. The firm down the road that does pick it up runs the same revenue with fewer people, or more revenue with the same team. You only see that difference later, in their rates, their margin, and in the waiting time they do not have to give new clients.
Companies structurally misjudge how fast something like this moves. In 1996, Dutch companies expected that by the end of 1998, 23 percent of them would have an internet connection. It turned out to be 42 percent (CBS). Not to scare you: to show why we work it out instead of claiming it.
The first step takes 30 minutes.
What the leak does in a year
€012 x the monthly leak
example, assumptions above
The firm down the road: same revenue, fewer people
Or: more revenue, same team
WinSphereAI and automation for firms and agencies with 10 to 100 people. First costed out, then built, then kept running.