by Konstantin Tskhay​

June 24th

Why HR Struggles with AI integration: 4 Barriers to Address

↓

Hi, Friends,

Last week, I wrapped up a fun panel discussion as part of the HR.com People Analytics Advisory Board. I sat down with Jeff Higgins, Helen Friedman, and Shane McGowan to discuss barriers to AI implementation in the organizations.

If you missed it, check it out here.

It was a timely conversation given many of my CHRO colleagues are currently tasked to push AI into their organization.

In fact...

Right now, almost every CHRO has been handed a clear, urgent mandate from their CEO and the Board: Integrate AI into the organization’s core and become an AI native (whatever that really means).

The pressure is massive.

Yet, despite all this pressure, most HR and Analytics teams are stuck.

Why?

Well, they are running into some gnarly limitations when it comes to AI transformation.

If you look under the hood of most corporate HR departments today, how are they actually using advanced generative AI?

They are

• drafting internal emails
• writing job descriptions
• summarizing basic policy documents

These use cases are fine.

But there are many, many, MANY more use cases to AI that we, as HR and People Analytics professionals talk about.

• Data at manager's fingertips
• Workforce planning
• Predictive analytics

These latter cases are much more complex and require us to address 4 analytics challenges:

1. Execution / Value Gap
2. Data Governance Gap
3. Talent Gap
4. Analytics Gap

Let's dive in.

Barrier 1: The Execution Gap

Let’s start with the ultimate trap in People Analytics and AI:

During our live panel session, we ran a live poll asking leaders to identify the single biggest barrier to People Analytics that, if completely solved, would unlock the greatest organizational impact.

You might assume the answer was "bad data" or "lack of executive buy-in."

However...

A staggering 57% of leaders pointed to the exact same problem: The Execution Gap.

So, what is the Execution Gap?

It is the reality where great insights are produced, but they are never translated into actual business decisions and actions.

Think about how a traditional People Analytics team operates. We hire brilliant data scientists. We buy expensive software. We aggregate millions of rows of data.

And then what do we do?

We hand our executives a complex dashboard with 50 different charts and expect them to miraculously figure out what to do with it. We assume that if we just provide the data, leadership will change their behavior.

They won't.

As Helen noted during our panel, leaders have their own personal experiences and biases. If you tell a leader something they already believe, they will nod and agree. But the moment you present AI-generated data that contradicts their gut feeling—or requires them to change their management behavior—you hit a massive wall.

They will immediately question the data. They will call the AI a "black box." They will refuse to execute.

The Solution:

Brining AI into the organization is all good.

However, before doing so, you need to ask:

AI for what?

What exactly am I trying to solve by brining this new model or tool into the organization. The challenge, of course, is not to get swayed by cool marketing materials, but to think deeply about why are we doing what we are doing.

Barrier 2: Data Infrastructure & Governance

You can drop a world-class AI model on top of garbage data.

But you shouldn't.

As we discussed heavily on the panel, our HR systems are highly fragmented and siloed. The ATS data doesn't often connect with the performance management data. Which sits completely separate from the financial ERP. Our teams are still not aligned on data definitions, data ownership, or data flows.

So, what happens when we force AI to answer complex business questions using inconsistent data?

We get hallucinations.

Let's look at standardization. We have seen organizations with over 150 different transaction codes just for terminating an employee!

When every single manager codes a departure differently, the data becomes useless. It reaches a point where neither your HR expert, your People Analytics leader, nor your AI can make any sense of the underlying themes.

But here is the dangerous difference between humans and machines.

A human analyst will look at the data, recognize that the data is a complete mess, and push back. AI will not. Instead of giving you deep insights, the AI will confidently make up quotes and hallucinate answers because it cannot process the chaos. In fact, it might add more chaos from it's world wide knowledge.

The Solution:

Before you invest millions into shiny AI tools and specialized teams, you have to fix your data.

I am hopeful that the intense pressure to adopt AI will force organizations to finally invest in their data governance initiatives.

In the AI era, Data Governance is no longer a "nice-to-have."

It is a non-negotiable.

Barrier 3: Specialized Analytics Expertise

We have a massive talent gap in HR right now.

Historically, an "HR Analyst" was someone who was highly proficient in Excel. They were a report builder. The CFO asked for a headcount report, and the analyst exported the CSV, ran a pivot table, and emailed it over.

AI completely obliterates that job description.

Today, AI can generate that headcount report in 4 seconds. The basic calculations are fully commoditized.

So, what happens to your People Analytics team?

They must evolve into specialized strategic advisors.

Your team now needs the technical skills to leverage the most advanced models, audit the algorithms for bias, validate the outputs, and—most importantly—translate complex data into a strategic business narrative.

If your People Analytics team lacks this specialized expertise, they will simply become passive administrators of the AI software. They won't know how to challenge the machine, and they won't know how to translate the machine's output to the CEO.

As an example, the other day, I tried using AI to run some more complex regression models, finding that the thing did not check for assumptions, did not specify proper standard errors, and used the wrong variance-covariance matrix.

Each time, it apologized to me for not doing something and mixing up the output.

So, from that standpoint, do you think you have an expert on your team to challenge AI appropriately and properly?

The Solution:

You must upskill your analytics team. Stop training them on how to pull reports. Start training them on statistical validation, prompt engineering, and executive storytelling.

It will pay dividends for you and for the entire company for a relatively small investment.

Barrier 4: Navigating Adoption of AI into Analytics

How should we actually be using AI now?

There is a massive push in the industry for "Democratization."

The idea is that we should give every front-line manager a ChatGPT-style interface linked to our workforce data, allowing them to ask complex questions without relying on an analyst.

But when we ran the live poll asking what the most valuable use of AI in People Analytics is today, the audience completely rejected that idea.

Only a fraction chose democratization or predictive modeling.

The overwhelming winner, at 36%, was Automation.

Specifically: Automating tedious data preparation, cleaning, and scheduled reporting.

This is the ultimate paradox of AI adoption. Everyone wants the sexy, predictive "Minority Report" models that tell us who is going to quit tomorrow. But the actual highest-ROI activity is incredibly boring pain point most of our teams face day to day.

Your People Analytics team is currently drowning in data work. They spend 90% of their week pulling data, cleaning formatting errors, and merging spreadsheets manually.

Only a tiny portion of the time goes to actually analyzing the data for the business.

The Solution:

Use AI to relentlessly automate the tedious data preparation and cleaning. Buy your team their time back. When your analysts spend 10% of their time prepping data and 90% of their time driving strategic change management, your Employee Lifetime Value (eLTV) will skyrocket.

The CHRO Playbook: Operationalizing AI

If you want to transition your People Analytics function from a "reporting factory" to a strategic business driver, here is your immediate roadmap:

1. Automate the Data Cleaning Work

Stop paying brilliant analysts to clean spreadsheets. Deploy AI immediately to handle the tedious data preparation, standardization, and scheduled reporting.

2. Lock in Your Data Governance

You cannot drive action if leadership is terrified of how the AI is using sensitive data. Audit your internal systems. Standardize your HR codes. Build the privacy guardrails first so you can deploy AI safely and securely.

Yes, data governance work can be very expensive. But if you don't do it, you are risking spending millions on AI that will force you back to data governance work later.

And at that point, it will be even more complicated.

3. Shift from "Reporting" to "Directing"

Stop giving leaders raw data and expecting them to interpret it. Your analytics team needs to take a definitive point of view. They must translate complex people data into actionable, clear insights that force a specific business decision.

4. Measure the Action, Not the Output

Focus on measuring the impact of AI. AI adoption as a concept is great only in as much as it solves business problems. So, if you cannot answer AI for what with a number, you must go back to the drawing board.

K

---

Whenever you’re ready, there are 2 ways I can help you:

#1

If you’re still looking to get started in People Analytics, I recommend starting with my affordable course:

​Practical People Analytics: Build data-driven HR programs to 10x your professional effectiveness, business impact, and career. This comprehensive course will teach you everything from building an HR dashboard for business results to driving growth through more advanced analytics (i.e., regression). Join your peers today! ​
​

#2

If you are looking for support in your human capital programs, such as engagement, retention, and compensation & benefits, and want to take a more data-driven approach, contact me at Tskhay & Associates for consulting services. Or simply reply to this email!