Faysal AzizAI Chat for Auditors
Using ChatGPT and Gemini with evidence and professional judgement
Practise AI-assisted audit prompting, data exploration, risk discussions and sampling-plan review using illustrative data and verified evidence.
Why this course
This one-day workshop explores how ChatGPT and Gemini can assist audit preparation, analysis and communication. Participants practise scoped prompts, inspect AI-generated summaries and discuss potential risks and tests using illustrative, non-confidential data.
AI suggestions are hypotheses or drafts, not audit evidence, authoritative accounting advice or independently valid sampling plans. Auditors retain responsibility for checking calculations, source documents, current standards and methodology. Tool access and data-processing features depend on the selected account and approved environment.
Learning outcomes
The workshop teaches participants to:
- Write and refine prompts for selected audit tasks and contexts.
- Review summaries, charts and suggested anomalies against the underlying data.
- Use AI suggestions to support, not replace, risk assessment and test-procedure design.
- Check technical or regulatory explanations against authoritative current sources.
- Critically review draft sampling approaches, assumptions, selection methods and calculations.
- Explain basic AutoML/anomaly-detection concepts and their limitations.
Prerequisites
- Basic understanding of auditing concepts and procedures
- Familiarity with data analysis tools (e.g., Excel)
- An open mind and willingness to explore AI's potential in auditing
A laptop and approved access to the selected chat tools. Exercises use prepared fictitious or appropriately anonymised data, not confidential client records.
7 modules
011 — Chat tools and audit boundaries3 topics
- Large language models and their applications; distinguish ChatGPT as a product from its underlying model.
- Compare selected ChatGPT and Gemini features, account access and data-handling requirements.
- Exercise: ask about an illustrative audit procedure, then verify the answer against the relevant source.
022 — Prompting for audit tasks4 topics
- Craft scoped prompts with context, assumptions, requested evidence and output format.
- Adapting prompts for different audit scenarios (financial statement audits, operational audits, compliance audits, etc.)
- Refine prompts and independently check the output; fluency is not accuracy.
- Exercise: draft prompts for fictitious transactions or internal-control scenarios.
033 — AI-assisted data exploration4 topics
- Summarise a prepared dataset using supported tool features and reconcile totals with spreadsheet calculations.
- Explore potential outliers or unusual patterns; anomalies are not proof of fraud or error.
- Generate and check visualisations, labels, calculations and data coverage.
- Exercise: examine an illustrative financial statement and verify each suggested area of concern.
044 — Risk focus and test drafts4 topics
- Generate candidate risk areas and red flags for the auditor to assess.
- Exploring hypothetical scenarios with ChatGPT to assess audit risks and control weaknesses
- Draft possible test procedures, then review suitability and evidence requirements.
- Exercise: simulate a risk discussion for an illustrative audit.
055 — Technical knowledge and source checking4 topics
- Use AI to suggest explanations or research questions, not substitute for authoritative industry knowledge.
- Locate current regulatory or standards sources and verify their applicability; AI output may be stale or incorrect.
- Query accounting/auditing concepts and cross-check with applicable authoritative guidance.
- Exercise: compare an AI explanation of a selected standard with the actual source and identify unsupported claims.
066 — Sampling-plan critique4 topics
- Draft and critique sampling approaches for a defined objective and population.
- Distinguish targeted item selection from statistical sampling and assess population completeness and selection bias.
- Check sample-size assumptions, sampling risk and calculations using approved audit methodology; a chat response does not establish statistical validity.
- Exercise: compare illustrative sampling scenarios and review the appropriateness of each approach.
077 — AutoML and anomaly detection overview4 topics
- Introduction to AutoML concepts and tools
- Demonstration of a prepared AutoML workflow suitable for the selected anomaly-detection task; model capabilities and data requirements vary.
- Review model outputs, false positives and follow-up investigation needs; predictions are not audit conclusions.
- Optional exercise with a small prepared dataset and an available no-code platform, time permitting.
A programme built around your team.
Share your training goals and requirements.