§ Playbook · HR
The AI playbook for HR
HR faces AI twice: as a user automating its own casework and hiring funnels, and as the function that must carry the people consequences of every other department's AI plan. This playbook covers both jobs - plus the legal exposure nobody budgets for and the 88% value gap nobody mentions at the vendor conference.
Free · no form, no follow-up sequence
Gartner found 61% of HR leaders in advanced stages of implementing AI - and 88% saying it has produced no significant business value. HR is currently running the largest gap between activity and outcome of any function. Closing it is the job.
88%
of HR leaders say AI has not delivered significant business value - while 61% are in advanced implementation.
Gartner, October 2025
1.1bn
rejected applications in scope in Mobley v. Workday, the AI-screening discrimination case.
US federal court filings, 2025
12d → 4d
Chipotle's time-to-hire with conversational AI; application completion up from 50% to 85%.
CNBC / Fortune, 2025-26
§ Part one
What actually changes.
The production evidence in HR clusters in two places. High-volume hiring: Chipotle's conversational assistant cut time-to-hire from twelve days to four and nearly doubled application completion. And case deflection: IBM's AskHR handles 94% of routine HR interactions across 11.5 million contacts a year - and IBM's total headcount went up, with the savings redeployed. The transactional layer of HR automates well. That was never the interesting part of the job anyway.
The second change is bigger and mostly missing from vendor decks: every other chapter of this series lands on HR's desk. Marketing's junior roles, IT's L1 tier, finance's graduate intake - someone has to redesign those roles, retrain those people and run those consultations properly. AI makes HR's admin smaller and HR's actual purpose larger. Few functions can say that.
01
High-volume hiring accelerates dramatically
Screening, scheduling and candidate conversation at hourly-hiring scale is the standout case - Chipotle, and Unilever's 50,000+ interview hours saved before it. Where volume is high and stakes per decision are low, the machine earns its keep.
02
HR casework deflects like IT tickets
Policy questions, letters, payroll queries, onboarding admin. IBM's 94% containment is the benchmark - built on years of knowledge and process work, not bought in a quarter. Deflection still is not resolution; check which one your dashboard shows.
03
Workforce planning gets real data
Skills inventories, attrition modelling, scenario planning that updates itself. The capability HR always claimed on the slide becomes genuinely buildable - if the people data underneath is clean enough to trust, which is its own project.
04
Recruiters move up the funnel
Sourcing and screening compress; relationship, assessment judgement and closing grow. The recruiter who is mostly a scheduler disappears; the one who can actually sell a role to a doubtful candidate becomes more valuable.
05
The judgement work stays human
Employee relations, grievances, restructures, the difficult conversation done properly. No serious evidence supports automating any of it, and your employment lawyers will happily confirm.
06
HR becomes the AI change function
Role redesign, consultation, retraining, honest communication - the people mechanics of everyone else's AI plan. This is HR's largest new workload and its strongest claim to the executive table in years.
§ Part two - the honest bit
What it costs you.
Start with the legal exposure, because it is the part vendors mention quietest. A federal court certified a nationwide age-discrimination collective over Workday's screening - filings reference over a billion rejected applications. The EEOC's first AI settlement cost iTutorGroup $365,000 for auto-rejecting older applicants. The EU AI Act classifies hiring systems as high-risk, with obligations from August 2026 and fines up to €15m or 3% of global turnover. Using the tool does not transfer the liability. You are the deployer; it is yours.
Then the internal truth: transactional HR shrinks. IBM's AskHR took over the work of a couple of hundred HR roles. Coordinators, schedulers and Tier-1 HR service desk are the exposed layer - the same honest conversation HR will supervise everywhere else, held first about itself. Consider it practice.
01
Transactional HR ops gets smaller
Shared-services queues, scheduling, letter production, first-line queries. The redeployment story is real - IBM grew - but it requires actually redesigning roles, not just deleting them and hoping.
02
The bias liability is yours, not the vendor's
Mobley, iTutorGroup and the EU AI Act all point one way: deployers carry it. Bias-test before deployment, audit on a schedule, document everything, and get the indemnities into the contract while you still have leverage.
03
Your HRIS is now also a risk surface
The decade-long platform investment is being defended with bolted-on agents and attacked in court simultaneously. Renewal conversations should include liability, audit access and bias-testing rights, not just modules.
04
Manager capability is the real bottleneck
Gartner: 8% of HR leaders think their managers can use AI effectively. Every workflow you automate lands on managers who cannot yet supervise it. Their capability is the gating investment, and nobody has budgeted it.
05
Candidates are gaming the funnel you automated
AI-written applications flood AI-run screening - machines interviewing machines while the humans wait outside. Volume metrics look splendid; signal quality quietly collapses. Design for it.
06
HR must eat its own cooking first
If HR automates its own admin with honest role redesign, consultation and retraining, it earns the standing to run everyone else's change. If it does it badly, everyone will know exactly how the rest will go.
§ Part three
What good looks like.
Good HR runs a smaller transactional core and a larger strategic one. Machines handle the queries and the scheduling; humans handle judgement, relationships and the hard conversations. Hiring systems are bias-tested and defensible. And HR owns the organisation's AI workforce transition - the role redesigns, the retraining, the honest communication - as its core work, not a side project.
01
A deflected core, honestly measured
Routine casework automated with resolution tracked, escalation clean, and the people data underneath governed well enough that the answers are actually right.
02
Hiring that would survive discovery
Every automated screening decision explainable, bias-tested on a schedule, documented, with human review where stakes are high. Compliant with the AI Act because it was designed that way, not retrofitted in a panic.
03
People on the people work
ER, coaching, restructures and consultation staffed with the hours the automation returned. The point of removing the admin was never fewer HR people - it was HR finally doing the job on the tin.
04
HR running the whole organisation's transition
A live map of which roles change function by function, retraining with budgets attached, consultation done properly, and an honest internal narrative. The workforce plan for the AI decade, owned by name.
The change, across all five pillars
People
HR redesigns its own roles first, then everyone else's - with the honesty it will demand of other functions.
Processes
Casework automated with clean escalation, and consultation run the way you would mark others doing it.
Data
People data governed well enough to trust the answers, the analytics and the audit trail it feeds.
Platforms
HRIS and screening tools bias-tested on a schedule, documented, and defensible in front of a regulator.
Partners
Vendor contracts with audit access and indemnities negotiated while you still have the leverage.
§ Part four
The first 90 days.
Two tracks at once: automate one piece of your own admin properly, and audit the legal exposure in the hiring stack you already run. One earns credibility; the other might save you a courtroom.
Ⅰ
Weeks 1-2
Audit what you already deployed
Every AI touchpoint in hiring and HR ops: what screens, what scores, what rejects, on what data. Map it against the EU AI Act and current litigation. This list is longer than you think.
Ⅱ
Weeks 3-6
Automate one queue, bias-test one funnel
Stand up automation on one high-volume casework or scheduling queue with resolution measured. In parallel, run an independent bias test on your highest-volume hiring funnel and fix what it finds.
Ⅲ
Weeks 7-10
Measure, document, redesign
Resolution and satisfaction on the queue; audit trail and remediation on the funnel. Redesign the affected HR roles openly - consultation done the way you would mark others doing it.
Ⅳ
Weeks 11-13
Take the bigger job
Publish the organisation-wide workforce transition map: which roles change where, what retraining exists, how consultation will run. Put HR's name on it. This is the seat at the table; sit in it.
The day-90 scorecard · what you should be able to show
§ Part five
The traps, so you can skip them.
Trap 01
Buying compliance from the vendor's brochure
The court cases name deployers. 'Our vendor assured us' is not a defence anyone has ever enjoyed giving under oath. Test it yourself, on a schedule, in writing.
Trap 02
Automating the empathy
A grievance handled by chatbot is a constructive-dismissal claim with a transcript. The moment a case turns human, a human takes it - visibly and fast.
Trap 03
The 88% club
Advanced implementation, no measurable value - the current HR default. Every deployment gets a baseline, an owner and a number, or it is a subscription with a lanyard.
Trap 04
Waiting on the AI Act
Deferral rumours are not a compliance strategy. Build to the high-risk requirements now; they are also just good practice with a legal deadline attached.
Trap 05
Skipping the manager problem
Automating workflows that untrained managers now supervise moves the risk downstream and calls it progress. The 8% number is the constraint; fund it.
Trap 06
Preaching change, dodging it at home
An HR function that automates everyone's roles but never redesigns its own has spent its credibility precisely where it needs it most.
§ Before you go
Six questions worth an honest hour.
Score yourself before someone with a rate card does. If more than two answers are uncomfortable, that is not a crisis - it is an agenda.
The self-assessment
§ Start with the workshop
Book your
Catalyst Workshop.
Bring the strategic change you have been circling. In 1-2 executive days, we will turn it into the first move, the case for it, and the mobilisation plan.