← All posts

AI in HR: What's Real, What's Coming, and How to Get Started

95% of organizations report no meaningful return on their AI investments. Here's why — and the practical steps to change it.

Here's a number that should make every HR leader uncomfortable: 95% of organizations report no meaningful return on their AI investments (MIT, "The GenAI Divide," 2025).

And here's what makes it truly strange: at the same time, more than half of HR professionals now use AI multiple times per day. In our own polls across 50skills webinars, 55% of HR attendees report using AI tools daily or more. Only 6% say "never."

So everyone's using it. Almost nobody's getting organizational value from it. What's going on?

The answer is deceptively simple: there's a massive gap between individual productivity and organizational capability. Using ChatGPT to draft a job description is useful. Having an AI agent that automatically screens every application, generates evaluation reports, routes qualified candidates forward, and notifies the hiring manager — all within a governed, auditable process — is transformational. Most organizations are stuck at step one.

Where We Are Today: The Sidekick Era

Let's be honest about what most HR teams are actually doing with AI right now:

  • Drafting job descriptions (1-3 hours manually to 10-20 minutes with AI)
  • Writing internal communications (newsletters, policy summaries)
  • Meeting summaries and action items (45-60 min per meeting to 5-10 min)
  • Employee handbook Q&A (5-10 min per query to instant)
  • Performance review prep (30-45 min per employee to 5-10 min)

This is genuinely valuable. The time savings are real — 80-90% on many tasks. But it's still the "sidekick" model: a human asks a question, an AI gives an answer, the human acts on it.

The problem is that most organizations stop here. They hand everyone a ChatGPT license and call it an AI strategy. According to our polls, only 20% of HR organizations have moved to actually automating workflows, and a mere 2% have built what we'd call a connected AI foundation.

The 95% who get no return? They're mostly stuck in sidekick mode.

What People Can Already Do — But Aren't

The technology for the next level already exists. It's here, and some organizations are already using it. The gap is awareness and implementation, not technology.

AI Inside Processes, Not Beside Them

The real shift happens when AI stops being a tool you go to and starts being embedded inside your actual work processes:

Candidate screening: When someone applies for a job, an AI agent immediately reads their CV and cover letter, scores them against the job requirements, generates an evaluation report, and routes qualified candidates forward — all automatically. The recruiter's job shifts from reading 200 CVs to reviewing 15 AI-evaluated candidates.

Expense reimbursement: An employee submits a receipt. An AI agent checks whether it's a valid receipt, verifies the amount matches the claim, determines the correct general ledger code, and either approves it or flags it for human review.

Photo validation: A company hiring foreign workers needs visa-compliant photos. An AI agent checks the photo against official requirements (background color, face angle, expression) and tells the employee immediately if they need to resubmit.

Incident reporting: An employee files an incident report. AI analyzes severity, routes it to the appropriate department, posts to the right Slack channel, and if severity is high, sends SMS alerts to key team members.

These aren't hypothetical scenarios. They're running in production, processing thousands of actions, saving measurable hours.

The Numbers From Real Customers

  • Festi (Iceland's largest retail group): 200+ hours of manual input avoided in their stock option program alone
  • Islandshotel: 80% reduction in time spent processing employee benefit applications
  • GRID: 90% less time spent on new employee onboarding
  • Icelandair: Presented to Iceland's entire HR community how they use 50skills for everything

The Bigger Transformation

Before we look ahead, it's worth understanding what's already happened in software development — because it shows us the pattern.

Developers no longer write most of their code by hand. They describe what they want to an AI agent — in plain language — and the agent writes it. The developer's job shifted from typing code to reviewing, directing, and deciding what to build next. A project that used to need a project manager, a designer, three programmers, and a QA tester can now be handled by one person working with an AI agent.

This has happened before. Two hundred years ago, 90% of the population farmed. Today it's 2% — and we produce more food than ever. When spreadsheets arrived in 1979, people thought they'd destroy the accounting profession. Instead, they created the field of financial analysis.

AI is following the same pattern. The question isn't whether this change is coming to HR. It's whether you're ready.

How to Get Started Today

If this all feels overwhelming, it shouldn't. The organizations seeing real value from AI in HR didn't start with a grand strategy. They started with one problem.

Step 1: Use Chat-Based AI Daily

If you're not already using ChatGPT, Claude, or Gemini for work every day, start there. Get comfortable. Learn what these tools are good at and where they make things up. Build intuition.

Step 2: Map Your Processes

Get a piece of paper and draw your processes. Not your dream processes — your actual ones. What happens when someone gets hired? When someone leaves? When an employee requests reimbursement?

Step 3: Pick Your First Process to Automate

Don't pick the flashiest one. Pick the one that wastes the most people's time. Usually it's onboarding (it touches everyone), employee requests (they're high-volume), or offboarding (nobody likes doing it, so it gets neglected).

Step 4: Use a Platform That Gives You Both Governance and AI

Don't choose between control and intelligence. The whole point of what's happening right now in HR tech is that you can have both.

Step 5: Iterate

Each process you automate frees time. Use that time to automate the next one. The flywheel effect is real.

Step 6: Upskill — Everyone

This isn't just a tools problem. It's a skills problem. Everyone in your organization needs to learn how to use AI in their daily work, understand how AI connects to the tools they already use, and redesign internal workflows for the AI era.

The bar for building is dropping. The bar for thinking is rising.

The Bottom Line

The organizations that will benefit most from AI in HR aren't the ones with the biggest technology budgets. They're the ones who start mapping their processes today.

AI is getting better every month. The models are more capable, the tools are more accessible, the costs are dropping. But AI can only help you if it has something structured to work with. A brilliant AI agent with no process to run is just another chatbot.

Map your processes. Start one. See what happens.