EXPOSED
This catch-all covers technicians who prep samples, run and calibrate instruments, collect field data, and log results for scientists — the modal worker spends real hours with hands on equipment and specimens, which robotics and AI cannot yet take. But the screen half of the job — transcribing readings, spreadsheet calculations, chart generation, QC flagging, drafting summary reports — is already automatable, and lab automation platforms are eating routine pipetting and plate handling. There is no licensure and no client relationship to defend the role, so the surviving tier is the person who can troubleshoot a misbehaving instrument and vouch for whether a result is real.
Mixed — a routine tier and a judgment tier. At 9 the split is roughly even: mounting specimens, swapping columns on a chromatograph, re-seating a leaking fitting and hauling a sampling kit into the field are still hand work, but the run logs, unit conversions, calibration-curve math, control-chart plotting and the boilerplate 'results within spec' memo are the other half of your week and LIMS-plus-liquid-handler setups already do them, which is why this sits below the 14+ band held by trades that never touch a screen.
Hands-on in uncontrolled environments. 13 reflects that a large share of these technicians work outside a controlled bench — surface-water and soil sampling, wildlife or vegetation plots, stack and ambient air monitoring, crop or greenhouse trials — where terrain, weather and uncooperative specimens defeat fixed automation; it is not higher because plenty in this catch-all spend their days at a fume hood or a bench instrument in a room with stable temperature and power.
No licence, no signature requirement. 4 is the ceiling for a role with no state licence and no personal signature: your data goes out under the PI's, lab director's, or certifying officer's name, and even in NELAP/ISO 17025 shops the accreditation attaches to the laboratory and its technical manager, not to you, so employer-specific training records are the only credential standing between you and replacement.
Anonymous artifact production. 4 is set by who reads your output — an internal scientist, a QA reviewer, or a regulator receiving a data package — none of whom chose you personally or would follow you to another employer; the informal credit you earn with a PI for clean, defensible runs is real but it never becomes the thing being purchased.
Meaningful discretion. 7 acknowledges genuine discretion inside written limits — deciding a calibration has drifted enough to rerun, rejecting a sample for broken chain-of-custody or hold-time exceedance, judging a peak from baseline noise — but SOPs, method manuals and a supervising scientist define the boundaries, and the consequential call on whether a result is reported or a study conclusion drawn is escalated, not yours.
Has AI actually changed your work?