Spikesz

I Explain AI, HR tech and EU rules in plain language for HR and People leaders who don’t have time for nonsense. No fluff, no shiny theory. HR is changing fast.

At SPIKESZ we help HR, Payroll and IT teams grow up fast in a world where AI, GDPR and multi-country complexity collide. Just the kind of expertise you only get after twenty years of rolling out HR systems across Europe, fixing the mess behind global templates, and helping leaders stay compliant when the rules shift overnight. We make AI in HR safer, cleaner and genuinely useful. We redesign proce

sses that people actually follow. And we prepare organisations for the new era of high-risk AI systems, audits and accountability. If you’re scaling across countries, integrating new tech, or trying to untangle the “we thought GDPR covered this” confusion, we’re the partner who has seen it, fixed it and knows where the real risks hide. Let’s make sure your systems, and your decisions, are ready for it.

04/02/2026

Your dashboard shows 95% adoption. Your executives think the implementation was a success.

But your HR team? They're still working evenings to manually reconcile data. Still keeping "backup" spreadsheets. Still double-checking everything before payroll runs.

People are logging into the system. But they're treating it like theater—performing the steps they're supposed to do, then going back to their spreadsheets to do the actual work.

I call this Theater Adoption.

Managers click "approve" in the system, then email HR to ask what they just approved. Department heads enter goals, then track the "real" goals in Excel. The official process happens...but nobody actually trusts it yet.

So they've built an entire shadow operation underneath. And it's exhausting.

This isn't a training problem. Your team knows HOW to use the system. They just don't trust that it actually works. So they're duplicating everything as a safety net.

The fix isn't more workshops. It's resetting the operating model so people can stop doing everything twice.

Anyone else seeing this gap between what your metrics show and what's actually happening?

03/02/2026

Why "More Training" Is a Dead End

Stop throwing training dollars at an authority problem.

I was reviewing a training budget last week. €80,000 for additional user workshops. The system had been live for four months.

"They keep asking the same questions," the HR leader said. "They must not understand how it works."

But when I dug in? Users knew exactly how to submit a time-off request. They just didn't trust it would actually get approved.

So they emailed their manager first. Then submitted it in the system. Then followed up with Gillian in Accounting to "make sure it went through."

Three steps where there should be one.

This isn't a training problem. It's an ownership problem.

When users know how to click the button but choose to email Gillian instead, you don't have a skill gap. You have an authority gap.

Can managers override the system? Can HR adjust things manually? Can Gillian just... fix it?

During implementation, those questions had answers. Post-go-live? That clarity vanished.

So users do what humans do when authority is unclear: they ask the person who always knew before. They create workarounds. They protect themselves.

More training won't fix this.

You need to reset who owns what. Document the decision rights. Clarify the boundaries. Make it crystal clear who can do what.

That's an Operating Model Reset, not another workshop.

Seen this pattern in your organization?

𝘊𝘰𝘮𝘮𝘦𝘯𝘵 "𝘵𝘦𝘮𝘱𝘭𝘢𝘵𝘦" 𝘣𝘦𝘭𝘰𝘸 𝘰𝘳 𝘴𝘦𝘯𝘥 𝘮𝘦 𝘢 𝘮𝘦𝘴𝘴𝘢𝘨𝘦 𝘪𝘧 𝘺𝘰𝘶 𝘸𝘢𝘯𝘵 𝘮𝘺 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯 𝘳𝘪𝘨𝘩𝘵𝘴 𝘮𝘢𝘱𝘱𝘪𝘯𝘨 𝘵𝘦𝘮𝘱𝘭𝘢𝘵𝘦—𝘪𝘵 𝘩𝘦𝘭𝘱𝘴 𝘥𝘪𝘢𝘨𝘯𝘰𝘴𝘦 𝘸𝘩𝘦𝘳𝘦 𝘺𝘰𝘶𝘳 𝘢𝘶𝘵𝘩𝘰𝘳𝘪𝘵𝘺 𝘨𝘢𝘱𝘴 𝘢𝘳𝘦 𝘩𝘪𝘥𝘪𝘯𝘨.

22/01/2026

To every CHRO holding an “AI policy” draft: Please, put it down.

A document asking employees to “use AI responsibly” is not governance. It is theatre.

I see organisations spending months wordsmithing policies, while their software quietly trains people to take shortcuts.

You cannot policy your way out of a bad design.

If the policy says “verify all outputs,” but the system produces a confident, perfectly formatted answer in two seconds, the software will win.

Every time.

The path of least resistance always does.

Real ethics don’t live in the handbook. They live in constraints.

The block that stops a high-risk action. The friction that forces a justification. The default that protects privacy even when someone forgets.

We design workflows that enforce compliance, instead of asking for it.

Most organisational ethics failures are design failures misattributed to people. Stop blaming users for the architecture you approved.

20/01/2026

AI is already in your HR stack. You just didn’t label it.

You might be blocking ChatGPT, but you should be aware that your vendors are shipping “intelligent features” by default.

The ATS ranks candidates. The engagement tool summarises sentiment. The HRIS suggests pay bands.

Each feature sounds small. Together, they form an ungoverned decision layer.

This is where the hidden exposure is.

Who owns the logic behind those suggestions? Usually, no one.

The vendor calls it “configurable” and hands you the risk. Procurement checks privacy clauses, not decision weight. HR Ops just keeps the tickets moving.

This is governance drift.

You didn’t decide to adopt risk. You accumulated it by default.

And when an employee asks, “Why was I ranked this way?” “The system suggested it” is not a legal defence.

I run forensic workflow audits to answer one question: Where is this system making decisions you cannot explain?

𝗧𝗵𝗲 𝗯𝗼𝘂𝗻𝗱𝗮𝗿𝘆: If you cannot produce an evidence trail without asking a vendor for help, you do not own your compliance.

19/01/2026

HR leaders think “human-in-the-loop” protects them. Well it doesn’t.

In most systems I audit, the human is not a safeguard. They are a liability buffer.

The operational reality is like this:

You deploy an AI tool to screen candidates or summarise feedback. You tell regulators a human reviews every decision.

But the interface tells a different story.

Approval is one click. Rejection is four. And the reviewer has 300 items to clear before lunch.

Under that pressure, the human stops reviewing. They start rubber-stamping.

The loop still exists on paper. The cognitive check is gone.

When bias shows up later, the audit trail blames the manager for “not paying attention.” That is false accountability.

If your system relies on human vigilance to prevent harm, you didn’t design a safeguard. You designed a scapegoat.

This is exactly what I simulate in a human-in-the-loop stress test. Not whether the control exists, but whether it still works when people are tired.

𝗧𝗵𝗲 𝗽𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲: Compliance is not a policy problem. It is a user experience problem.

15/01/2026

We mapped a global payroll system for a mid-sized client in 3 weeks.

Before we started, the HR Director told me: "We are pretty sure we know where all our data is."

𝗧𝗵𝗲 𝗥𝗲𝘀𝘂𝗹𝘁: We found 4 AI tools they didn't know they were using.

1. A chatbot on the careers page recording conversations.
2. A scheduling tool reading calendar metadata.
3. An "optimization" feature in their payroll software sending data to the US.
4. A legacy CV parser that hadn't been updated since 2019.

"Pretty sure" is not a compliance strategy.

You cannot govern what you cannot see.

If you haven't updated your data map in the last 6 months, you are likely operating blind. Let's turn the lights back on.
__

Enjoy this? ♻️ Repost it to your network and follow 𝗦𝗣𝗜𝗞𝗘𝗦𝗭 for more.

14/01/2026

I don't work in AI Governance because I love reading 200-page EU regulations. (I actually hate them).

I do this because I know what it feels like to be a number in a system.

Years ago, before I was a consultant, I saw a colleague let go based on "performance metrics" that were clearly flawed. The data said they were underperforming. The reality was they were fixing everyone else's mistakes, which the system didn't track.

The machine was efficient. But the machine was wrong.

Now that we are giving AI the power to screen resumes and track productivity, that risk is a thousand times higher.

I fight for compliance because compliance is just a fancy word for fairness.

We aren't protecting data. We are protecting the people behind the data.
__

Enjoy this? ♻️ Repost it to your network and follow 𝗦𝗣𝗜𝗞𝗘𝗦𝗭 for more.

13/01/2026

I audited a "secure" HR tech stack last week.

Here is what the vendor brochure promised: "GDPR Compliant & Private."

Here is what I found in the fine print of their Terms of Service update:

⚠️ Data Training: They reserve the right to use "anonymized" employee data to train their public models.
⚠️ Sub-processors: They send data to a third-party server in a non-adequacy country for "optimization."
⚠️ Retention: They keep rejected candidate CVs for 3 years (default setting) even if you only asked for 6 months.

This is what I call "Shadow AI."

You think you bought a closed system. But the default settings are wide open.

Go check your vendor contracts today. Search for the word "Training."
__

Enjoy this? ♻️ Repost it to your network and follow 𝗦𝗣𝗜𝗞𝗘𝗦𝗭 for more.

12/01/2026

To the Project Lead who is terrified of the Works Council meeting tomorrow:

Stop trying to "sell" them on AI.

When you walk in there with a slide deck full of "efficiency gains" and "speed," you are speaking the wrong language. They don't care about speed. They care about safety.

I have sat in dozens of these meetings. The ones that go well always follow this structure:

1. 𝗔𝗰𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝘁𝗵𝗲 𝗥𝗶𝘀𝗸 First Start by listing the dangers before they do. "We know this tool could be biased."

2. 𝗦𝗵𝗼𝘄 𝘁𝗵𝗲 "𝗛𝘂𝗺𝗮𝗻 𝗶𝗻 𝘁𝗵𝗲 𝗟𝗼𝗼𝗽" Prove that a human, not the algorithm, pushes the final button.

3. 𝗧𝗵𝗲 𝗘𝘅𝗶𝘁 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 Explain how you will turn the tool off if it breaks.

If you treat the Works Council as an obstacle, they will become one. If you treat them as Quality Assurance, they will save you from a compliance disaster.
__

Enjoy this? ♻️ Repost it to your network and follow 𝗦𝗣𝗜𝗞𝗘𝗦𝗭 for more.

05/01/2026

Most HR Directors think "Transparency" means adding a one-line disclaimer to a job ad.

"We use AI in our hiring process."

I have bad news. Under the upcoming EU standards, that line is useless.

Transparency isn't about confessing that you use a tool. It is about explaining how the tool makes decisions.

If a candidate asks why they were rejected, and your answer is "The computer said you weren't a fit," you are opening the door to a lawsuit.

Here is the difference between Legal Compliance and Actual Transparency:

❌ Legal Minimum: "We use automated decision-making." ✅ HR Trust: "Our AI scans for keywords X, Y, and Z. It does not analyze your photo, age, or address."

The first protects the company (barely). The second respects the human.

Stop writing disclaimers. Start writing explanations.

31/12/2025

𝗔 𝗻𝗲𝘄 𝗱𝗮𝘁𝗮 𝗽𝗿𝗶𝘃𝗮𝗰𝘆 𝗿𝘂𝗹𝗶𝗻𝗴: 𝟯 𝗛𝗥 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀 (𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗜’𝗱 𝗱𝗼 𝗠𝗼𝗻𝗱𝗮𝘆 𝗺𝗼𝗿𝗻𝗶𝗻𝗴)

New privacy rulings rarely sound like “HR news”.

Then two months later, HR is asked:
“Can you prove this decision was fair, and that the data use was lawful?”

My 3 HR takeaways whenever a fresh ruling touches worker data:

1. 𝗣𝘂𝗿𝗽𝗼𝘀𝗲 𝗰𝗿𝗲𝗲𝗽 𝗶𝘀 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗿𝗶𝘀𝗸.
Data collected for HR admin quietly gets reused for scoring, monitoring, or “risk” prediction.

2. 𝗧𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝗰𝘆 𝗶𝘀 𝗻𝗼𝘁 𝗮 𝗽𝗼𝘀𝘁𝗲𝗿.
If employees would be surprised by the processing, you’re already in the danger zone.

3. 𝗩𝗲𝗻𝗱𝗼𝗿𝘀 𝗱𝗼 𝗻𝗼𝘁 𝗰𝗮𝗿𝗿𝘆 𝘆𝗼𝘂𝗿 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆.
If your tool ranks people, HR owns the story you can defend to a regulator, works council, or judge.

𝗪𝗵𝗮𝘁 𝗜’𝗱 𝗱𝗼 𝗠𝗼𝗻𝗱𝗮𝘆 𝗺𝗼𝗿𝗻𝗶𝗻𝗴:
Send one email to your top 3 HR vendors: “List all AI/ML features active in our tenant, what data they use, and what human controls exist.”

𝗣𝗦: 𝗜𝗳 𝘆𝗼𝘂 𝗮𝘀𝗸𝗲𝗱 𝘁𝗵𝗮𝘁 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝘁𝗼𝗱𝗮𝘆, 𝘄𝗵𝗶𝗰𝗵 𝘃𝗲𝗻𝗱𝗼𝗿 𝘄𝗼𝘂𝗹𝗱 𝗴𝗶𝘃𝗲 𝘁𝗵𝗲 𝘃𝗮𝗴𝘂𝗲𝘀𝘁 𝗮𝗻𝘀𝘄𝗲𝗿?

30/12/2025

Stop buying AI for HR.

Not because AI is bad.
Because you’re about to automate a process you cannot explain.

Here’s the uncomfortable truth from implementations:
If your workflow lives partly in Excel, email, and “whoever remembers”… AI will amplify the mess.

𝗗𝗼 𝘁𝗵𝗶𝘀 𝗳𝗶𝗿𝘀𝘁, 𝗶𝗻 𝟰𝟱 𝗺𝗶𝗻𝘂𝘁𝗲𝘀:
1. Pick ONE decision flow (e.g., shortlist → interview → offer).
2. List every system and data source that touches it (ATS, HRIS, assessment, spreadsheets).
3. Mark where “scoring”, “ranking”, or “recommendations” happen.

If you cannot draw it on one page, you are not ready for “smart features”.
You are ready for a map.

𝗣𝗦: 𝗪𝗵𝗮𝘁’𝘀 𝘆𝗼𝘂𝗿 𝗺𝗲𝘀𝘀𝗶𝗲𝘀𝘁 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗳𝗹𝗼𝘄 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄? 𝗛𝗶𝗿𝗶𝗻𝗴, 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲, 𝗽𝗮𝘆, 𝗼𝗿 𝗶𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗺𝗼𝗯𝗶𝗹𝗶𝘁𝘆?

Adres

Amsterdam

Website

https://www.spikesz.com/

Meldingen

Wees de eerste die het weet en laat ons u een e-mail sturen wanneer Spikesz nieuws en promoties plaatst. Uw e-mailadres wordt niet voor andere doeleinden gebruikt en u kunt zich op elk gewenst moment afmelden.

Snelkoppelingen

Wilt u dat uw bedrijf hét hoogst genoteerde Bedrijf in Amsterdam wordt?