The AI in HR Practitioner Gap: What I Actually Built With the Claude API (And Why Most HR Leaders Are Still Just Watching)

There is a version of AI in HR that lives in keynote decks and LinkedIn carousels. Someone standing on a stage telling you that AI is going to transform everything, that the future of work is here, that you need to get ready.

Then there is the version I work in. Southeast Wisconsin. Real organizations. Real constraints. Real humans whose livelihoods depend on whether the People function gets it right.

I am Drew Soule. I use a wheelchair. I was born with a physical disability. And I have spent 15 years building HR functions in hypergrowth tech, aerospace, financial services, and healthcare, in rooms that most HR practitioners never see. That background matters when I tell you what I am about to tell you about AI.

I am not here to hype it. I am here to show you what an HR automation practitioner actually does with it.

Why the AI in HR Practitioner Gap Is Real and Getting Wider

The gap in this space is not a knowledge gap. Most HR leaders understand what AI is. The gap is a practitioner gap. Very few people in this field are actually building with it, testing it, failing with it, and iterating on it in the context of real HR work.

I started building custom HR workflow automations using the Claude API because I had a specific problem. I was running HRBP work supporting an 800-person global Product and Engineering organization with 10 to 40 concurrent employee relations cases every week. Not 10 to 40 cases per quarter. Per week. That volume is not something you manage with good intentions and a spreadsheet. You need systems. You need structure. You need to eliminate the mechanical load so your judgment is available for the situations that actually require it.

So I built the systems. And I want to talk about what that actually looks like as a generative AI HR use case, because most of what gets published about AI in HR is either so theoretical it has no operational value, or so superficial it tells you nothing about the real work.

What Claude API HR Workflows Look Like in Practice

Let me be specific, because specific beats general every time.

When I am working an ER case, there are layers. There is the documentation layer, the pattern recognition layer, the communication drafting layer, and the judgment layer. For a long time, all of those layers demanded the same thing: my direct attention and time. The result was that genuinely complex situations, the ones that required hard calls and careful thinking, were competing for cognitive space with tasks like drafting an initial acknowledgment letter or structuring a case summary from scattered notes.

The Claude API changes that calculus. I built workflows that handle the mechanical drafting load. Initial case documentation structure. First-draft communication templates calibrated to tone and situation type. Summary scaffolding from raw interview notes. None of this replaces my judgment. All of it clears the path for my judgment to operate where it matters.

This is what HR analytics AI integration looks like when it is grounded in operational reality. Not a dashboard that tells you what already happened. A workflow that changes how the work gets done while it is happening.

I want to say something clearly here because I think it gets lost in most AI in HR conversations. AI is not magic. The Claude API is a tool. A powerful, genuinely useful tool. But the value it creates is entirely dependent on whether the person building with it understands the work first. I understood the work. That is why the tool produced something useful. The tool did not make me better at HR. My HR experience made the tool useful.

The 53% Problem and What Structured Thinking Actually Looks Like

When I joined a healthcare organization and built an entire People Programs function in 60 days, one of the first things I did was diagnose why people were leaving. The answer was not in a model. It was in pattern recognition across qualitative exit interview data, tenure data, and manager-level cohort analysis.

Fifty-three percent of exits were happening in the first 90 days. That number is not a statistic. That is a signal. It tells you exactly where the break is in the employment experience. It tells you that the organization had a hiring story that was not matching the onboarding reality. It tells you that managers were not equipped to close the gap between expectation and experience in the window that mattered most.

We redesigned onboarding. We built manager-level accountability structures. We measured a 22% attrition reduction and over $400,000 in cost avoidance.

I tell that story because it is the right frame for thinking about AI in employee relations and AI in People Operations broadly. The value of AI is not that it finds answers for you. The value is that it accelerates the structured thinking that lets you find answers faster, with more consistency, at higher volume. You still have to know what questions to ask. You still have to know what the patterns mean. The judgment is yours. The load-bearing mechanical work does not have to be.

AI Ethics in People Operations: The Questions I Am Actually Asking

I want to spend some time here because AI ethics in people operations is an area where I see a lot of performative concern and not a lot of rigorous thinking.

I use a wheelchair. I have navigated every element of a professional career in a world that was not designed with me in mind. I have experienced firsthand what it means when systems, processes, and structures carry embedded assumptions that exclude people without ever intending to. That experience is not incidental to how I think about AI in HR. It is central to it.

The risk in AI-assisted HR work is not that the tool will do something dramatic and obviously wrong. The risk is that it will do something subtly wrong in a way that is consistent and scalable. Bias that is invisible in a one-off human decision becomes systemic when it is baked into an automated workflow running at volume.

So when I build, I ask hard questions. What assumptions are embedded in this template? Who does this communication style work for, and who does it leave behind? What does this pattern recognition miss? Who is not represented in the data I am feeding this process?

Those are not abstract ethics questions. They are operational design questions. And if you are an HR automation practitioner building real systems, they are non-negotiable.

## What Labor Relations Taught Me That AI Cannot Replace

Before I was building with the Claude API, I was sitting across the table from union representatives in a 1,200-person multi-site manufacturing and supply chain organization with active collective bargaining agreements. That work has almost nothing in common with the AI-enhanced ER workflows I run now, and it has everything in common with them.

What labor relations teaches you is that the human relationship is the actual product. Not the contract. Not the process. The relationship. And the relationship is built or broken by whether the people across the table believe you are operating in good faith, whether you understand their reality, and whether your accountability matches your stated values.

AI in HR lives or dies by the same standard. The technology is not the point. The trust is the point. When I use AI to draft faster, document more consistently, or identify patterns at scale, the value to the people I serve is that I show up to the actual conversation with more clarity, more preparation, and more genuine attention for them. That is the only version of AI in HR I am interested in building.

HR Tech Practitioner Wisconsin: Why Location Is Not Incidental

I want to say something about being an HR tech practitioner in Wisconsin that I do not see discussed enough.

The organizations I work with here are not Bay Area startups with venture capital runways and a culture of experimentation. They are manufacturers, healthcare systems, financial services firms, and growth-stage tech companies operating in a Midwestern market where trust is built slowly and credibility has to be earned every single time. The tolerance for theoretical frameworks is low. The demand for practical results is high.

That is not a limitation. It is a crucible. It means that everything I have built, every workflow, every framework, every AI-assisted process, has been tested against the hardest standard: does it work for real people in real organizations with real constraints?

That is the standard I hold AI in HR to. Not does it look impressive in a demo. Does it hold up under pressure, at volume, for people whose jobs and livelihoods are the stakes?

The answer, when you build it right, is yes.

Closing: Build It or Watch It

I said at the top that there are two versions of AI in HR. The version that lives in keynotes and slide decks. And the version I work in.

The gap between them is not going to close on its own. It closes when practitioners stop waiting for the technology to be explained to them and start building with it. When HR leaders treat AI as a craft, not a feature. When the standard for success is not whether the tool is impressive but whether the people it serves are better off.

I am an AI in HR practitioner based in Southeast Wisconsin. I use a wheelchair. I have been in the rooms where the hard calls get made. And I am building systems that are worth building.

Not because AI is magic. Because the humans building it understood the work first.