AI in HR: 2025 Lessons from Google to McDonald’s

A lot of HR teams talk about AI like it’s a tool you “add” to hiring, performance, or employee engagement.

Most companies get this wrong. The most useful AI in HR isn’t a feature you bolt onto an old process—it’s what happens when you redesign the process so people, data, and decision rights actually match how work gets done.

That’s the through-line I keep hearing from the strongest HR leaders featured in UNLEASH’s 2025 interviews—across Google, McDonald’s, SAP, HUGO BOSS, Moderna, and others. Different industries, same message: AI is forcing HR and workforce management to become more systems-minded, more business-aligned, and more disciplined about governance.

This post is part of our “AI in Human Resources & Workforce Management” series, and it translates those 2025 executive perspectives into practical moves you can make in 2026—especially if your goal is better workforce planning, faster talent acquisition, cleaner performance analytics, and employee engagement that doesn’t feel like surveillance.

The real shift: HR is becoming an “operating system” for work

Answer first: AI works in HR when HR owns the operating model—not just the tools.

One of the most telling 2025 signals came from Moderna’s decision to merge HR and Digital/IT under one executive umbrella. The point wasn’t re-org theater. It was about fixing a problem nearly every enterprise has: HR designs “people programs,” IT designs “systems,” and then everyone wonders why the employee experience feels stitched together.

When people strategy and technology strategy are managed as one system, three AI use cases get dramatically easier:

• AI-enabled workflow design: You can automate the right steps because you understand the end-to-end flow of work (not just the HR slice).
• Workforce planning and org scaling: You can connect headcount, skills, project demand, and productivity signals without fighting data ownership battles.
• Employee experience consistency: You stop asking employees to navigate five portals and a dozen policies to get one thing done.

Here’s what I’ve found in practice: if HR can’t map how decisions are made (who decides, with what data, and under what policy), AI will amplify the mess. If HR can map that, AI becomes a force multiplier.

What to do in Q1 2026

Run a short “flow of work” audit on one process that crosses HR + business + IT:

1. Pick one high-friction journey (internal mobility, frontline scheduling, performance check-ins, or onboarding).
2. Map the workflow steps and identify where people hand off work or re-enter the same data.
3. Mark the points where judgment is required vs. where rules apply.
4. Only then decide where AI should lead, support, or step aside.

That last line matters. The best HR leaders in 2025 weren’t arguing for more AI everywhere—they were arguing for smarter placement of automation.

Leadership is changing because AI changed the pace of work

Answer first: AI doesn’t just automate tasks—it changes what “good leadership” looks like.

Google’s talent and learning leadership described it well: we’re not just in constant change; we’re in a change of eras. That’s not a slogan. It’s an operating reality for HR because:

• Skills become outdated faster.
• Teams re-form more often.
• Work is increasingly done through platforms and cross-functional systems.

In this environment, AI in HR and workforce management needs to support leaders in two concrete ways:

1) Faster, clearer decisions (without removing accountability)

Leaders are drowning in signals—engagement scores, productivity metrics, hiring funnel stats, performance ratings, attrition risk models. AI can help synthesize and prioritize, but it cannot be the “decider” for sensitive people outcomes.

A practical standard to use:

• AI summarizes: “Here are the top drivers of attrition for this team in the last 90 days.”
• Leader decides: “Which driver do we address first, and how?”
• HR governs: “What decisions are allowed, what evidence is required, and what’s audited?”

2) New manager capabilities (because old playbooks won’t scale)

If you’re rolling out AI copilots, managers need training that’s more specific than “use it ethically.”

Teach three skills:

• Prompting for management work: feedback drafts, coaching plans, interview guides.
• Data interpretation: understanding confidence levels, bias risk, and when a model is unreliable.
• Escalation judgment: knowing when to involve HR, legal, or employee relations.

If you want measurable impact, define it upfront: shorten time-to-fill, improve internal mobility, reduce regrettable attrition, increase manager quality scores. AI programs that don’t declare a scoreboard tend to become novelty.

HR credibility in 2026 will come from governance, not enthusiasm

Answer first: The organizations that win with AI in HR will be the ones that can explain, audit, and defend their people decisions.

SAP’s HR leadership put a stake in the ground: no function besides HR can guide companies through workforce transformation successfully.

I agree—and not because HR is “closest to people.” It’s because HR is the only function positioned to create decision governance across the employee lifecycle:

• What data can be used in hiring?
• Which attributes are off-limits?
• How do you validate a selection model?
• What’s the appeal process?
• Who signs off on automation?

This matters because regulators, works councils, and employees are paying attention. In 2025, the conversation shifted from “Can we use AI?” to “Can we prove it’s fair, explainable, and secure?”

A simple governance model that actually works

If you’re building AI into talent acquisition or performance analytics, use a three-layer control system:

1. Policy layer: definitions (what’s allowed), documentation standards, retention rules.
2. Model layer: validation, bias testing, drift monitoring, periodic re-approval.
3. Process layer: human review steps, exception handling, employee transparency.

If you can’t describe these in one page, you’re not ready for AI-driven decisions at scale.

Frontline and global workforces: AI succeeds when local teams can adapt

Answer first: In global workforce management, the best AI approach is “central standards, local experimentation.”

McDonald’s people leadership talked about empowering regions to experiment, share learnings, and scale what works. That’s exactly how AI should be deployed in large, distributed workforces.

A centralized HR team can define the guardrails:

• Data definitions (what counts as turnover, absence, productivity)
• Core skills taxonomy
• Approved AI vendors and security requirements
• Bias and compliance testing

But local leaders must have room to adapt workflows because frontline operations aren’t identical across countries, brands, or labor markets.

Where AI helps frontline workforces immediately

In practical workforce management terms, AI can deliver value fast in:

• Demand-driven scheduling: predicting staffing needs and reducing over/under coverage.
• Faster hiring for hourly roles: screening support, interview scheduling, candidate Q&A.
• Training personalization: role-based microlearning triggered by performance signals.
• Retention interventions: identifying churn patterns early (then letting humans act).

One warning: if the AI outcome is “We predicted you’ll quit,” employees will feel watched. If the outcome is “Your manager got better tools to fix scheduling fairness and training support,” employees feel supported. Same dataset, different intent.

High performance without a toxic culture: keep human judgment by design

Answer first: The safest way to use AI in employee engagement and performance is to preserve human judgment at the moments that matter.

Heineken’s culture message is the one HR analytics teams should print and hang up: don’t give up human judgment—context and compassion are non-negotiable.

AI can support performance management, but only if you design it to:

• Reduce admin work: draft summaries, consolidate feedback, surface themes.
• Improve consistency: remind managers about standards, calibration guidance.
• Protect fairness: flag anomalies (for example, unusually low ratings for one demographic group).

AI should not be used to generate final ratings, determine pay outcomes, or replace difficult conversations. If your implementation encourages that, you’re building a future employee relations problem.

A better pattern for AI-assisted performance

Use AI as a “pre-meeting assistant,” not a “judgment engine.”

• Before check-ins: AI compiles accomplishments, peer feedback themes, goal progress.
• During: manager and employee discuss context, constraints, and growth.
• After: AI helps write the summary and development plan—clearly labeled as a draft.

That approach gets you efficiency without removing the human part that employees actually care about.

HR as a business function: stop leading with tools, lead with outcomes

Answer first: AI in HR drives leads (and internal buy-in) when you pitch it as a business outcome with measurable workforce metrics.

HUGO BOSS’s HR stance—business first, HR second—lands because it forces discipline. If you’re trying to get budget for AI in talent acquisition or workforce planning, don’t say:

• “We need an AI platform.”

Say:

• “We’re losing revenue because time-to-fill for critical roles is 18 days too long, and our hiring managers spend 6+ hours per week on low-value screening steps. We’re redesigning the process and using automation where it’s safe.”

Then show the measurement plan.

A practical ROI checklist for AI in HR

If your team wants to build a credible business case in 2026, track:

• Time-to-fill and time-to-productivity (hiring + onboarding)
• Quality of hire (90/180-day outcomes, not vibes)
• Internal mobility rate (and time to move)
• Manager effectiveness metrics (retention, engagement, team performance)
• Schedule quality for frontline teams (fairness, stability, coverage)

Tie AI investment to one or two of these, not ten.

What HR leaders should do next (especially if you need results in 90 days)

AI in HR and workforce management is heading into a more serious phase in 2026: fewer experiments for experimentation’s sake, more pressure to prove ROI, and more scrutiny around responsible AI.

If you want traction fast, start here:

1. Pick one workflow that creates measurable pain (hiring throughput, scheduling, internal mobility, performance admin).
2. Redesign the process first, then insert AI where it clearly saves time or improves decision quality.
3. Create governance you can explain—policy, model, and process layers.
4. Train managers for real usage, not theoretical ethics slides.

The leaders highlighted across 2025 UNLEASH interviews weren’t selling magic. They were building systems: tighter partnerships between HR and tech, clearer leadership expectations, stronger governance, and more employee-centered design.

If you’re planning your 2026 roadmap now, here’s the question worth ending on: Which single people decision—hiring, scheduling, mobility, performance, or retention—would improve the most if your data, workflow, and accountability finally matched up?