Team Wheel

HR Without Humans? Try It, And See You In Court

blog post August

HR Without Humans? Try It, And See You In Court

The Financial Times asked a simple question this week: does HR still need humans? The short answer is yes. The longer answer is that AI will automate a large share of HR’s repetitive work, but the organisations that win will be the ones that redesign HR around augmentation, accountability, and trust. Anything else risks legal blowback, poor decisions, and a talent brand you cannot fix with a chatbot.

What the hype gets right, and where it goes wrong 

The hype is not baseless. IBM’s internal “AskHR” has automated a big block of routine requests and processes, freeing people to focus on higher-value work. IBM says AskHR now automates more than 80 common HR processes, and external reporting around IBM suggests automation covers the vast majority of routine HR tasks.

Banks and large enterprises are also signalling real workforce shifts as AI scales. Public commentary around JPMorgan, for example, points to headcount reductions in some operations over the next five years as efficiency rises. Whether the final number is 5, 10, or 15 per cent, the direction of travel is clear.

Where the hype goes wrong is assuming that fewer HR administrators means you can outsource judgement. You cannot. The moment you use AI to screen CVs, rank candidates, or trigger adverse decisions, you move into a regulated zone with real legal exposure. In Europe, recruitment, promotion and termination are classed as “high-risk” uses under the EU AI Act, which drags you into obligations around data quality, transparency, human oversight and post-market monitoring.

And the case law is forming. In the closely watched Mobley v Workday litigation, a US court recently allowed parts of an age-discrimination claim to proceed as a conditional collective action, spotlighting algorithmic screening risk whether you build or buy. Regardless of outcome, the direction is plain: if your system disadvantages a protected group, you had better be able to explain and correct it.

What we have been saying on LinkedIn Live 

On Team Wheel’s recent LinkedIn Live with Adrian Bingham, we discussed exactly this: AI will not replace HR, it will augment roles, change the shape of the team, and expose weak processes that need redesign. HR’s centre of gravity moves from transactional execution to product management of talent workflows, data stewardship, and change leadership. 

Keep an eye out for our upcoming LinkedIn Live with Simon Tetley where we will dig into AI in Talent Acquisition and Talent Management. We will cover sourcing copilots, interview agents, skills graphs, and the governance guardrails that make these safe and effective. 

The emerging playbook: design for augmentation 

1) Start with the work, not the tool. Map tasks, decisions and data flows, then decide which steps are automated, assisted, or human-only. McKinsey’s latest workplace research shows most companies are still rewiring processes to capture value, and that leadership behaviours, not employee resistance, are now the binding constraint.

2) Keep a human in the loop for people-critical decisions. For hiring, promotion and exit, require human review and create an auditable trail. This is both good practice and aligned with the EU AI Act’s expectations for high-risk use.

3) Invest in explainability you can actually use. If a model downgrades a candidate with a non-linear career path, your reviewer should see, and challenge, the features driving that score. The live legal risk around automated screening is a strong incentive to get this right.

4) Data governance is a people strategy. Bad data equals bad decisions. HR needs ownership of reference data, consent, retention and lineage. Evidence from the OECD’s multi-country studies suggests outcomes are better when workers are trained and consulted through adoption.

5) Measure more than cost-out. IBM’s lesson is useful: remove routine drudge, then redeploy capacity into skills, culture, and design of work. That is where value shows up beyond the first year.

6) Pilot with transparency and time-boxing. Pick one high-volume, low-risk flow, publish the success criteria, and time-box the pilot. MIT Sloan’s recent work on “vibe analytics” shows how cross-functional teams can interrogate data with GenAI and collapse analysis cycles from months to minutes. Use that speed to learn, not to bypass governance.

What this means for HR roles 

The jobs do not vanish, they evolve. 

  • Business-facing HR becomes product managers of talent journeys, with accountability for outcomes and ethics. 
  • People analytics becomes a revenue-relevant function, not a reporting factory. 
  • TA moves from requisition processing to market intelligence, skills mapping, and experience design, with AI handling triage and scheduling at scale. 

McKinsey’s global survey data suggests HR is among the functions most likely to report cost reductions from GenAI already, but enterprise-level EBIT impact is still maturing. Translation: the efficiency is real, but the strategic upside depends on redesigning work, not just buying tools.

Buyer beware: three traps to avoid 

  1. Proxy discrimination. If your historical hires are narrow, your model will be too. Bake in fairness testing, representative training data, and counterfactual checks. The Workday case is a warning sign for everyone, not a verdict on any one vendor.
  2. Shadow automation. Chatbots quietly making decisions without policy approval is how you end up on the front page. Declare where automation starts and stops, and log it. 
  3. Tech without trust. Employees will adopt AI faster when they are trained, consulted, and can see recourse routes for appeal. That is not just soft change management, it shows up in the OECD evidence base.

So, does HR still need humans? 

Yes. Remove the drudge work, absolutely. But keep humans where stakes and context are high, and turn HR into the function that designs safe, data-rich, human-centred workflows. That is how you get productivity gains without reputational or regulatory risk. The FT’s provocation is useful, but the better question is: how quickly can HR become the operating system for responsible AI at work.

 © 2025 Team Wheel. All rights reserved. This document and its contents are the intellectual property of Team Wheel. Reproduction, distribution, or use of any part of this document without the express written consent of Team Wheel is strictly prohibited.