Original LinkedIn edition published
These are original companion notes. Read the full newsletter on LinkedIn
The short answer
If AI capability affects pay or promotion, employers need a defined expectation, a fair chance to develop and demonstrate it, and a reviewable record of the evidence used. Tool use or course completion alone cannot establish proficiency or job impact.
Make the evidence match the pay decision
A survey reported by Reuters on 14 October 2026 suggests some financial firms are beginning to connect AI proficiency with raises, bonuses or promotion. The public account does not establish a shared definition of proficiency or a valid formula for reward. A policy should therefore specify what capability matters for each decision, why it matters and which evidence could reasonably demonstrate it before managers are asked to rate people.
Separate foundational knowledge, opportunity to use approved tools, quality of work and observed outcomes. The Sova-delivered Core-6 is the in-market generic knowledge assessment; it can inform a baseline or development discussion, but it does not measure performance in a particular job or determine pay. Evidence from work needs its own criteria, quality checks and human judgement. More prompts, logins or faster output can reflect access and task mix rather than better work.
Give employees clear expectations and comparable opportunities to learn and apply AI within their roles. Where approved tools, data permissions or useful tasks differ, document those constraints before comparing people. Calibrate decisions across managers, retain a rationale, allow challenge and revisit whether the criteria produce unequal or misleading results. CLEAR in the edition is an employer discussion framework, separate from the Core-6 assessment; it is not a scored product module.
Illustrative example: a reward review across different access
A service manager considers two analysts for a progression award. One has an approved AI tool for routine summaries and can show fewer corrected errors in reviewed work. The other handles restricted records and has no approved tool for that task, but has shown sound knowledge and improved a manual quality check. The manager records the different opportunities, examines verified work quality against each role's agreed criteria, and avoids using prompt volume as a common score. A calibration panel checks the rationale before a decision is final.
Questions to take into your next meeting
- Write down the capability expected for the role and the specific reward decision it informs.
- Separate foundational knowledge from tool access, work quality and verified outcomes.
- Check whether employees had comparable approved tools, relevant tasks and development time.
- Record and calibrate the human judgement, with a route to correct or challenge it.
Put the idea to work
Audit one proposed AI proficiency criterion
Take one draft pay or promotion criterion. Ask what it claims to measure, what evidence a manager would collect, who lacks a fair opportunity to demonstrate it, and how a disputed rating would be reviewed. If the answer is mainly tool activity or self-reported confidence, revise the criterion before it affects reward.
Could you explain and defend the evidence behind an AI-related pay decision for two employees with different access to tools?
Sources and further context
- Reuters, 14 October 2026 Report on a Selby Jennings survey of finance executives about linking AI proficiency to reward; the reported survey does not provide a universal assessment standard
- European Commission, 14 October 2026 Survey of European SMEs and midcaps on AI and algorithmic management, including performance monitoring and trust in staff-related decisions
Core-6 measures common AI knowledge through Sova. The activities discussed here are separate ways to explore work and behaviour.
What evidence belongs in an AI literacy programme? Try the separate A/A/H and PRIME worksheet See what Core-6 measures