Writing
Notes on people, work and AI. Short, with a point of view and sources included.
Essays
AI won’t just take junior jobs. It can make juniors operate like pros.
AI gives every junior a patient bench of advisors and gives seniors leverage on their judgment. Whether that builds careers or deletes them is a design choice.
Abundance won’t distribute itself
Whatever happens to Musk’s 2036 prediction, how AI-driven abundance reaches people and work depends on decisions leaders make.
What the next HR service model should deliver
Design the next HR service model around the services employees need, one owner per journey and a 90-day pilot that earns the next decision.
Starting a career when AI can do the first draft
New entrants should use AI seriously and make their learning visible, while employers protect a real route to expertise.
Experts should use AI too. Linus Torvalds does.
Torvalds using AI on kernel work shows that expertise is what makes AI useful, with direct implications for how HR runs AI programs.
Short notes
Every employee move needs one owner
Start an HR redesign with the service and give each employee journey one accountable owner.
A policy bot is only as good as the policy behind it
Reliable AI answers in HR depend on owned, current policy knowledge and an assistant that knows when to stop.
Make it easy to reach the right person
Self-service should make HR easier to reach, with a simple route to a specialist when a case turns sensitive.
A closed ticket isn’t a solved problem
HR service metrics should show what happened to the employee, not just whether a ticket closed on time.
Start AI with the people who answer the questions
Test an internal assistant with HR service agents before putting AI in front of every employee.
Authority before access
Define what an HR agent may change before it gets access, and make sure a human reviewer can actually say no.
Agree what good looks like before the interview
AI can help prepare interviews, but only a clear, structured hiring brief makes the assessment fair and valid.
Tell candidates where AI is involved
Candidates deserve a specific, accurate explanation of where AI is used in recruiting and who to contact with concerns.
Workforce decisions start with the work
An AI exposure score can’t tell leaders what a role should deliver; mapping the team’s actual work can.
A skills profile needs evidence
Skills data becomes useful for talent decisions when it is tied to observable evidence of work.
Train for decisions, then make room to apply them
AI literacy means sound decisions in real work, and new skills only stick with time, an assignment and feedback.
Internal mobility runs on release conditions
A talent marketplace only works when leaders agree capacity, time and release conditions for internal moves.
Onboarding is done when the new colleague can contribute
Design onboarding backward from a first real contribution, with the manager owning expectations.
Ask the team while the decision is still open
Involve the people doing the work before AI and workflow decisions are locked, and close the loop afterward.
Help managers rehearse difficult conversations
AI role-play can help managers prepare for hard conversations, within clear limits on data and advice.
Check whether the tool improved the working day
Judge AI tools by the whole working day, since time saved on one task can vanish into review and extra volume.
Be clear about what shared time is for
Whatever the hybrid arrangement, teams get more from time together when they agree what it is for.
Decide what saved time is for before rollout
AI time savings only become value when the use of freed capacity is agreed upfront and hidden costs are counted.
Build a workforce plan you can adjust
Plan the workforce across adoption scenarios with clear triggers, instead of defending a single long-term number.
Scheduling software needs an owner too
Systems that allocate or evaluate work need a named owner and a correction route, whether or not they are called AI.
Start people analytics with the decision
Agree the business decision and the available actions before commissioning a predictive model.
Show us where the tool struggles
Evaluate HR technology on hard cases and running costs, with every affected function in one evaluation.