Cases

What we built, and what it delivers for you

What we built, and what it delivers for you

What we built, and what it delivers for you

WinSphere is the AI and automation partner for B2B service firms, founded by Rafael Matabadal. We determine which systems you need and in what order, and build them: foundation first (clean data, connected systems), then the workflows that cost your team time. Below are six systems we built for recruitment and search agencies, accounting firms and other B2B service firms of 10 to 50 FTE: from a sourcing agent that turns a vacancy into a screened shortlist in minutes, to a compliance check that is already running live today.

WinSphere is the AI and automation partner for B2B service firms, founded by Rafael Matabadal. We determine which systems you need and in what order, and build them: foundation first (clean data, connected systems), then the workflows that cost your team time. Below are six systems we built for recruitment and search agencies, accounting firms and other B2B service firms of 10 to 50 FTE: from a sourcing agent that turns a vacancy into a screened shortlist in minutes, to a compliance check that is already running live today.

Specification
App
Live

Live, today

We do not just build, it runs

We do not just build, it runs

We do not just build, it runs

A compliance check we built for a recruitment agency, already running live today.

A compliance check we built for a recruitment agency, already running live today.

Recruitment / compliance check · live

14 runs · real numbers, running live

Form in, substantiated status report out

How it works · schematic
Form
completed
Risk score
calculated automatically
PDF report
generated
Email + CRM
qualified lead

What we built.
A freelancer compliance check, built for a recruitment agency. The freelancer fills in the check and automatically receives a risk score plus PDF report by email, which they can use to substantiate towards the Dutch tax authority that they meet the requirements. A substantiation, not a legally binding judgment.

Outcome.
Running live: 14 successful runs up to and including 2 July 2026. Real run numbers, not an estimate.

14 runs live

through 2 July 2026

What it makes possible.
The same pattern, form → score → report, can be deployed for any intake-heavy process: handling more volume without the next hire.

Stack: n8n + document AI + CRM (live)

Recruitment

More placements from your existing candidates

More placements from your existing candidates

More placements from your existing candidates

Three systems for recruitment and search agencies, from sourcing to follow-up. Every block reads the same: situation, what we built, what it delivers.

Three systems for recruitment and search agencies, from sourcing to follow-up. Every block reads the same: situation, what we built, what it delivers.

Recruitment · running in production

From vacancy to screened shortlist, in minutes

How it works · schematic
Vacancy
one completed form
LinkedIn scan
up to 50 profiles
Score
role · sector · region
CRM
screened shortlist

Current way of working.
At an executive search agency, every assignment started with manual sourcing: endlessly scrolling through LinkedIn, assessing profiles one by one.

Problem.
Hours per vacancy: time lost to searching instead of speaking to candidates and advising clients.

What we built.
A sourcing agent turns one vacancy form into a search query, retrieves up to 50 profiles and scores them; only those who clear the threshold make it into the CRM.

Outcome.
From manual sourcing to a shortlist in minutes. Running in production at the agency.

indicative: ~2.5 hours → minutes

running in production

What it makes possible.
More vacancies at once, with the same team: the recruiter starts from a pre-selected list. More placements per recruiter, not more recruiters.

Stack: n8n + CRM + sourcing

Recruitment · built

From application to a phone call within 30 minutes

How it works · schematic
Application
comes in
Score A-E
with gap analysis
SMS + email
to the top candidates
Call < 30 min
recruiter alert

Current way of working.
Applications come in, the recruiter reads them, compares them with the vacancy requirements and has to follow up with the best ones quickly before they drop off.

Problem.
Manual screening takes time, and top candidates expect speed. Anyone who only hears back after two days is often already in talks elsewhere.

What we built.
A chain that scores every application against the requirements (A-E in the CRM), immediately sends the best ones a text + email and alerts the recruiter: call within 30 minutes.

Outcome.
Manual CV assessment per applicant disappears; top candidates are contacted in minutes, not days.

indicative: ~10-15 min per CV eliminated

top candidates in minutes, not days

What it makes possible.
Processing more applications and reaching the best ones faster, with the same team. No more top candidates lost to slow follow-up.

Stack: n8n + ATS + SMS/email

Recruitment · built + tested

Warm replies that no longer go cold

How it works · schematic
Reply in
in the inbox
Intent score
warm · lukewarm · no
CRM opportunity
warm rises to the top
Follow-up
alert or automatic

Current way of working.
Replies to LinkedIn and outreach land in the inbox. The recruiter has to spot the warm ones themselves, remember them and follow up.

Problem.
No system underneath. A strong reply left sitting on a busy day goes cold, and that is a missed placement.

What we built.
A layer that captures every reply, scores the intent with AI (warm / lukewarm / no), adds it as an opportunity in the CRM, alerts on warm and follows up itself.

Outcome.
In testing the system scored and routed correctly: a test reply received a score of 84 = warm and landed at the top. No warm reply slips through anymore.

score 84 = warm (test)

indicative ~1.5 hours/week less triage

What it makes possible.
Recruiters focus on the candidates who are already there. More conversations from the same stream of replies, without an extra recruiter.

Stack: n8n + CRM + outreach tools

Custom software & dashboards

More than workflows: hosted apps on your own data

More than workflows: hosted apps on your own data

More than workflows: hosted apps on your own data

Besides automations we build custom software: apps with their own interface, login, database and dashboard, at a price that fits SMEs.

Besides automations we build custom software: apps with their own interface, login, database and dashboard, at a price that fits SMEs.

Recruitment · designed

Your own database, searchable by meaning

How it works · schematic
Vacancy text
paste it in
Semantic search
on your own database
Top 10
by meaning, with explanation

Current way of working.
An executive search agency in the cultural sector opened LinkedIn for every vacancy. Its own database with thousands of paid profiles remained untouched.

Problem.
The ATS searches by keywords. Someone with ten years of museum experience does not surface for “cultural policy advisor”, even though that experience is right there.

What we designed.
A hosted search app on their own database: paste the vacancy text and get a top 10 by meaning, not by keyword, each with an explanation.

Outcome.
From hours of LinkedIn sourcing to a shortlist in ±1 minute. Candidates you never found by keyword now do surface.

indicative: hours → ±1 min

top 10 by meaning

What it makes possible.
More placements from the database you already paid for. One extra placement from your own file already pays for the tool.

Stack: own data + semantic search (designed)

B2B · custom software

Custom work, built for a paying client

An intel dashboard that reads public market data

Intel-dashboard
Schematic view
Grootste spelers in de markt
Player A
Player B
Player C
Player D
Openbare marktdataautomatisch uitgelezen en samengevat

Current way of working.
A B2B company manually researched who the biggest players in a market were, document by document.

Problem.
Hundreds of hours of research, and by the time the overview was ready, it was already outdated.

What we built.
A secured dashboard that automatically reads public market data and summarises who the biggest players are, with its own interface, login and database.

Outcome.
Hundreds of hours of manual research, reduced to the push of a button.

indicative: hundreds of hours → 1 button

public market data

What it makes possible.
Think of internal dashboards, intel tools and semantic search apps on your own data. Properly secured, data for you only.

Stack: Next.js + Supabase + n8n + AI

How we keep it honest

Why there are no names and hard numbers here (yet)

We do not put client names or logos on display that we are not allowed to use, and we do not make up results. What you see above are systems we actually built, anonymised on request.

Numbers marked “indicative” or “indicative estimate” are substantiated estimates, not measured client results. The freelancer status check is the exception: it runs with real run numbers.

As soon as a client gives permission, a case with a name, quote and hard number will replace one of these scenarios.

Anonymised on request: no client names or logos. Numbers marked “indicative” are substantiated estimates, not measured client results, except for the freelancer status check, which runs with real run numbers.

Ready to automate your operation?

Ready to automate your operation?

Ready to automate your operation?

You get me on the line, not an account manager. Rafael

© 2026 WinSphere · winsphere.ai · KvK 93325614Den Haag, NL
WinSphere