people

Feedback and reviews

• SBI for every piece of feedback: situation, behavior (observable, not character), impact. "You're careless" is an accusation; "the last three releases shipped without the checklist run, causing two rollbacks" is feedback.
• Continuous > annual: nothing in a review may be a surprise — if it is, the manager failed earlier. Reviews summarize known feedback.
• Ratings need anchors and calibration across managers, or they measure manager generosity, not performance.
• Self-review + manager review; peer input for seniors. Recency bias is the default failure — require examples spanning the whole period.

Hard conversations

Structure: state the purpose in sentence one → the evidence (SBI) → pause and listen genuinely → the ask or decision → next steps in writing. Never sandwich serious messages in praise — it buries the signal and reads as manipulation. Kind and clear beats nice and vague ("clear is kind" — Brown).

PIPs — fairness test first

A PIP is legitimate only if: expectations were clear beforehand, feedback was given and documented, support was offered, and success is genuinely achievable. A PIP designed to fail is a termination without honesty — say so and route to counsel instead. Structure: specific gaps with evidence, measurable success criteria, support provided, weekly check-ins, 30–90 days, outcomes stated up front. Lawyer review before delivery, always.

Surveys

Pulse: 5–10 questions max, same core items over time (trend beats snapshot), one open text. Anonymity threshold ≥5 respondents per slice. The iron rule: never ask what you won't act on — an ignored survey costs more trust than no survey.

Policy drafting

Plain language, the why included, edge cases acknowledged, owner and review date stamped. Draft from the country's mainstream pattern (websearch), flag every legally-variable clause (leave, overtime, termination, monitoring, pay) for counsel. A policy nobody can find or read is not a policy.

Law in motion (2026 — flag and verify, never rule)

• EU Pay Transparency Directive (2023/970): transposition was due 7 June 2026; only a handful of states had fully transposed by autumn 2026, so national rules vary. Core duties: objective, gender-neutral pay criteria; employee right to pay information by category; gap reporting starting June 2027 (250+ employees yearly, 150-249 every three years, 100-149 from 2031); joint pay assessment when an unexplained gap of 5% or more persists. Build job architecture and pay bands now.
• UK Employment Rights Act 2025 (Royal Assent December 2025, phased): SSP from day one since April 2026; tribunal claim window six months from October 2026; unfair-dismissal qualifying period drops to six months from 1 January 2027 — probation reviews and documentation matter from month one.
• US non-competes: the FTC ban was vacated and the FTC dropped its appeals in September 2025; it now challenges non-competes case by case. State law governs, and several states ban most of them — don't draft one without counsel.
• AI in people decisions: under the EU AI Act, AI used for promotion, termination, task allocation, or performance monitoring is high-risk (obligations postponed to 2 December 2027 by the 2026 Digital Omnibus), and emotion recognition at work has been banned since February 2025. US rules include Illinois (notice and anti-discrimination for AI in employment decisions, from January 2026) and NYC Local Law 144 (bias audits for automated hiring and promotion tools). Keep a human accountable for every people decision.