Publications on AI for internal auditing practices are few and far between, and therefore deserve a mention. In Repères 2024 (Editions La Découverte), a PhD student and an associate professor at Paris-Dauphine University, a partner at Grant Thornton France, Hayk Hovhannisyan, Beatrice Bon Michelet, Nicolas Gasnier-Duparc, have written a report on the subject of:
“The impact of AI on the internal audit process”.
Based on a quantitative study carried out in 2022 and involving over 120 auditors, they investigate how new AI technologies are likely to profoundly change the field of internal auditing. As the authors point out, the AI studied here, “goes far beyond the ‘data analytics’ now used in both internal and external audit functions”.
Still limited use of AI, but high expectations
The first observation is blunt: the use of AI is still underdeveloped in this business area.
- Only 25% of respondents make frequent use of AI
- Only 11% of respondents regularly use Machine Learning
On the whole, however, auditors are certain that AI will enable them to:
- Reduce processing time (90% of respondents).
- Abandon sampling techniques in favor of a certain level of exhaustiveness (88% of respondents).
Other improvements, such as making the results obtained more reliable, are also mentioned.
But not without risks and difficulties
Of course, as with any new technology, the use of AI is not without risks:
- Parameterization errors (79% of respondents).
- Biases in data selection (80% of respondents).
- Information overload (52% of respondents).and loss of control in understanding the results provided by the system (64% of respondents).
And the difficulties are linked to the way in which data is organized within organizations:
- A sometimes substantial initial investment in data architecture (73% of respondents).
- Difficult access to databases (69% of respondents).
- Lack of IT knowledge among auditors (65% of respondents).
- The cost of developing or purchasing solutions (63% of respondents).
- Lack of database quality or reliability (48% of respondents).
- The time required to operate these solutions (45% of respondents).
- The ability of auditors to properly analyze results (33% of respondents).
- Lack of motivation (33% of respondents)
However, in a society where risks are increasingly numerous, and where it is important to be able to move from “what happened” to “what could happen” (Prokofieva 2022), AI seems to offer a real qualitative leap forward. Within a conservative area accustomed to slow progress, that also needs to capitalize on the results of previous audits.
But as the authors remind us, AI adoption won’t happen without major changes in methodology. These changes include:
- The need to collaborate with other company departments, all the more difficult for individuals accustomed to extreme confidentiality.
- Learning new skills in data-science and data-analysis.
- The ability to interpret the results produced by AI.
Finally, and this is true in all areas of the company, we need to learn to collaborate with the machine, to accept its errors, but above all to take advantage of them in feedback loops, in which the listener observes a machine error, and provides the machine with this example so that it doesn’t make any further mistakes.
See also: Kairntech for contract analysis and Kairntech RAG assistants