AI field guide

Use it responsibly · Guide 10 of 10

AI in your field

Find where AI helps in your work, and the checks your field still requires.

What it is great for in your field

Students

A well-designed tutor can help you learn more in less time, as long as it makes you do the thinking. 12

Show prompt

Try this

Quiz me on [topic] from my notes: [paste]. One question at a time. Explain what I got wrong, and finish with what to revise.

Then check Follow your course's AI rules, and test yourself with the tool closed.

Developers

Use it to explore code, write tests and draft small changes. The Code with AI guide has the full method. 34

Show prompt

Try this

Review this change as a strict senior engineer: [paste diff]. List bugs and security problems first, then anything that would be hard to maintain.

Then check Run the tests, and read every change before you merge it.

Health

In two small trials, AI scribes cut time on clinical notes for one product and improved burnout measures. Patient outcomes were not measured. 5

Show prompt

Try this

Using this made-up case for practice: [case], draft a clear clinical note and flag anything missing or unclear.

Then check Scribes made occasional significant errors, so a clinician checks every note. Keep real patient data out of unapproved tools.

Law

AI can help you understand and summarise documents, as long as you confirm every authority yourself. 67

Show prompt

Try this

Summarise this judgment for a law student: [paste]. Give the facts, the issue and the reasoning, and quote the paragraph behind each point.

Then check Find every case in an official database. One researcher lists 2,241 court decisions discussing AI fakes.

Finance

Use it to explain concepts and draft analysis. Firms reportedly use AI widely, and regulators reportedly apply their existing rules. 8

Show prompt

Try this

Explain [concept] to someone new to finance, with a worked example using simple numbers. Show each calculation so I can check it.

Then check Recalculate every figure yourself. Keep client data out of unapproved tools.

How to do it

  1. Find the rules for your field

    Courses, employers and regulators each set their own.

  2. Start with low-stakes work

    Practise on summaries, drafts and made-up cases.

  3. Check what your field would check

    Citations in law, figures in finance, notes in health, tests in code.

  4. Keep the final call

    Your name is on the work, so the decision stays with you. 9

Where to be careful

  • It can invent believable sources

    A lawyer was sanctioned in 2023 over fabricated cases cited after using ChatGPT. 6

  • Good evidence is still thin

    Few studies here measure real outcomes. No finance outcome study was found, and the health trials did not measure patient results. 58

Behind this guide

Notes, evidence and sources

Why we say this

4 sources, each with its catch.

NEJM AI · Late 2025

AI scribes can cut note time

Across two small trials, one product cut time on notes and burnout measures improved, with occasional clinically significant note inaccuracies. 5

Patient outcomes were not measured, and the papers were not read.

Search summary only: re-check at the source
Full detail

Two trials in NEJM AI (late 2025) tested ambient AI scribes for clinical notes. A UCLA trial (Lukac et al.) of 238 physicians compared two products with usual care and found one reduced time spent on notes, with modest burnout improvements and occasional clinically significant note inaccuracies. A pragmatic trial (Afshar et al.) of 66 practitioners reported work exhaustion down 0.44 on a 5-point scale (PubMed; News-Medical summary).

Small trials measuring time and burnout, not patient outcomes. Funding and vendor ties were not checked, citation details vary, and the papers were not read.

Strength: Moderate, small

Mata v. Avianca · 22 June 2023

Fake cases, real sanctions

A 2023 sanctions order concerned a lawyer who used ChatGPT and cited fabricated cases. 6

The court order itself could not be opened.

Search summary only: re-check at the source
Full detail

In Mata v. Avianca (S.D.N.Y.), a sanctions order of 22 June 2023 by Judge Castel concerned a lawyer who used ChatGPT and cited fabricated cases (overview; order).

The court order itself could not be opened (access refused), so details such as the sanction amount are not given.

Strength: Strong that it happened

AI Hallucination Cases database · Updated 29 October 2026

2,241 court decisions discuss AI fakes

One researcher's database lists 2,241 court decisions discussing hallucinated content. 7

Some are only alleged, and it is not a rate per filing.

Source page read
Full detail

Damien Charlotin's AI Hallucination Cases database listed "2241 cases identified so far" (updated 29 October 2026): 1,294 involving self-represented litigants, 884 lawyers and 33 judges. By country: USA 1,514, Canada 238, Australia 112, UK 89.

It counts court decisions discussing hallucinated content, some only alleged, and "does not track the (necessarily wider) universe of all fake citations". It is curated by one researcher and is not a rate per filing.

Strength: Strong that it happens and is rising; not a rate

Bank of England and FCA · 21 November 2024

Most surveyed firms reportedly use AI

A 2024 survey with 118 respondents reportedly found 75% of firms using AI, and the FCA reportedly says it has no AI-specific rules. 8

No outcome study was found, and no primary document was read.

Search summary only: re-check at the source
Full detail

A Bank of England and FCA survey (21 November 2024, 118 respondents) reportedly found 75% of firms using AI and a further 10% planning to within three years. The FCA says it has no AI-specific rules and supervises under existing frameworks, and FINRA Regulatory Notice 24-09 (June 2024) says existing supervision and communications rules apply to AI-generated content (law-firm summary).

Regulatory position only: no outcome study for finance was found, no 2025 survey was found, and no primary document was read.

Strength: Weak to moderate

Terms used in this guide
Ambient AI scribe
A tool that listens to a medical appointment and drafts the notes.
Hallucination
When an AI tool states something false as if it were true, such as a court case that does not exist.
Outcome study
Research measuring real end results, such as patient health, rather than time saved.
All 9 sources for this guide

  1. 1. Kestin et al. · June 2025

    Kestin et al., 2025 · full article

    Source page partly readStrength: Moderate

  2. 2. Maths students in Turkey (Bastani et al.)

    PNAS abstract · working paper

    Source page partly readStrength: Moderate

  3. 3. Peng et al. · February 2023

    randomised experiment

    Source page readStrength: Moderate
    • Vendor or interested-party source

  4. 4. Veracode 2025 GenAI Code Security Report · 30 July 2025

    Veracode 2025 GenAI Code Security Report

    Source page readStrength: Moderate
    • Vendor or interested-party source

  5. 5. Ambient AI scribes in NEJM AI · Late 2025

    PubMed · News-Medical summary

    Search summary only: re-check at the sourceStrength: Moderate, small

  6. 6. Mata v. Avianca · 22 June 2023

    overview · order

    Search summary only: re-check at the sourceStrength: Strong that it happened

  7. 7. AI Hallucination Cases database · Updated 29 October 2026

    AI Hallucination Cases database

    Source page readStrength: Strong that it happens and is rising; not a rate

  8. 8. Finance: survey and regulators · 21 November 2024

    law-firm summary

    Search summary only: re-check at the sourceStrength: Weak to moderate

  9. 9. UK Government AI Playbook · 10 February 2025

    UK Government AI Playbook

    Source page readStrength: Strong as official guidance