Science and Discovery lab: how AI is changing the tasks

Key facts

• In the ILO's 2025 index, medical records technicians score 0.52 out of 1 for exposure to generative AI, while nursing professionals score 0.25: in health, the paperwork is far more exposed than the bedside.^[1]
• Field roles such as environmental health inspectors (0.24) and agricultural technicians (0.26) are rated "not exposed".^[1]
• Employers in the WEF's 2025 survey listed big data specialists among the fastest-growing jobs and expect nursing professionals to grow significantly by 2030.^[2]
• Life sciences made up 6.4% of the Claude conversations Anthropic studied in February 2025.^[3]

In the Science and Discovery lab you test, measure, care and find things out: in a hospital lab, a factory's quality lab, a pharmacy, a river or a spreadsheet. This brief shows how AI is changing some of the tasks in that work, using the "tasks, not job titles" method: look at what people do, then ask which parts AI can help with.

How to read the evidence

Most scores below come from the International Labour Organization's 2025 index. It rates each occupation from 0 to 1 by how much of its current tasks generative AI could potentially do, then groups occupations from "not exposed" up to gradient 4, the most exposed.^[1] Three warnings:

• It measures potential, not what has happened. The ILO stresses that exposure does not mean a whole occupation will be automated.^[1]
• It covers generative AI only, not lab robots or new diagnostic machines. "Not exposed" is not a promise that a job will stay the same.
• Matching a lab task to an ILO occupation is our choice and could be argued. The "how sure" column is our judgement too.

Example tasks and how AI is changing them

Notice the pattern: the records and data side scores much higher than the hands-on side. That does not make any task "safe" or doomed. New equipment, rules or costs could change any of them, and these studies do not model Malaysian hospitals or labs.

Never trust an AI number unchecked

Science work lives or dies on numbers you can stand behind. When an AI tool works out an average, a dose, a concentration or a total, redo the sum yourself or check it against your own data. When it suggests a method, ask what it forgot to keep the same. When it cites a study, check that the study exists and says what the AI claims. Your name goes on the result, not the AI's.

Business, enterprise and AI run through every task

• Costing the work: pricing a set of lab tests or budgeting a small field study. The ILO's experts judged structured money estimates, such as a household budget, to be a task GenAI can do much of.^[1] You still check the prices and assumptions.
• Pitching a finding: explaining a result to a client, a clinic manager or a funder, including how sure you are. AI can help you prepare the slides; you answer the hard questions.
• Using AI tools safely: never put patient details or unpublished data into a public AI tool, and check every fact against the original source. UNESCO's AI framework for students places the ethics of AI and a human-centred mindset alongside AI techniques.^[4]

In the taster

In the Science and Discovery taster you run a fair test for a client, ask AI for a method and for your averages, and check both. A quality analyst does the same with any tool: what changed, what stayed the same, and would I put my name on this result?

For facilitators

Ask the lab practitioner to correct the table before learners use it. In the career-area tasters, health-science learners use the records, data and checking-results rows; environment-sustainability learners use the field sampling and compliance reporting rows. A health practitioner should read the nurse, pharmacist and lab rows before use.

What this means for you

• Hands-on lab, field and care work is among the least exposed to generative AI, on the ILO's measure. Records and data work is among the most exposed.
• Checking is a core skill. Fair tests, repeated readings and checking AI sums are what employers in testing and health will trust you for.
• Data skills cut both ways. Employers expect data jobs to grow, and data tasks are also among the most exposed to AI. Knowing what the numbers mean, and whether to trust them, is the part to get good at.
• Privacy is part of the job. In health and research you handle data that must never go into a public AI tool.

Your next step

1. Choose one task from the table and ask the lab practitioner: "Which part of this do you trust AI with, and which part never?"
2. Read how AI may change your career area: health and science, environment and sustainability or technology and AI. Then X-ray one job with the Tasks, not job titles method.