With some help
from AI 😊 we’ve made a
summary of issues to be addressed, to avoid mistakes in using artificial
intelligence and to implement it in such a way that the organisation and employees
profit from it.
Apart from the
more technical aspects – relevant for decision makers, compliance officers and
staff responsible for implementing and running the AI engines – there’s the question
about what the ‘average’ client-facing, operational or supporting staff member
needs to know and understand about artificial intelligence in general and the
AI used in the organisation specifically.
That’s where
i-KYC comes in with our e-learnings like the “AI Awareness in the Financial Sector”
course. The training provides an overview of artificial intelligence, its
application in financial services and focuses on what employees in various roles
need to be aware of when dealing with AI.
Apart from the risks outlined in the rest of
this article staff awareness on the topic of AI is crucial and all-staff training
is a must to raise that awareness.
The biggest
pitfalls of using AI in financial services revolve around algorithmic bias,
which could cause discrimination in decisions on lending, pricing or client
acceptance and potentially creates a lack of transparency if "black
box" models are used, preventing understanding how decisions are made. Bear
in mind that the model might not at all be hidden but might be defined in such
a complex way that operational or compliance staff do not get a good grasp of
the decision parameters.
Other major
risks include data privacy breaches, high development costs, regulatory
compliance hurdles and too many staff just not knowing enough about the risks
and impact of the use of AI in the organisation.
As indicated
in the introduction, we have listed the main risks and pitfalls in more detail
below.
1. Algorithmic
Bias and Discrimination
-
Perpetuating
Inequality: AI models trained on historical data can inherit and even amplify
existing biases, leading to discriminatory outcomes in lending, insurance, and
risk management.
-
Hidden
Bias – only found once results after implementation are analysed
2. "Black
Box" Opacity and Lack of Transparency
-
Unexplainable
Decisions: Many advanced AI systems act as "black boxes," making
it difficult to understand, explain and justify decisions, which is critical
for audits and regulatory compliance.
-
Erosion
of Trust: A 2024 survey found that 89% of financial firms cited the lack
of transparency as the main barrier to AI adoption, as staff find it harder to
trust non-transparent systems and explain outcomes to customers.
3. Data
Privacy and Security Vulnerabilities
-
High-Value
Targets: AI systems process vast amounts of sensitive financial data,
making them a potential target for cybercriminals.
-
Data
Poisoning: training data can be manipulated or altered, potentially compromising
AI models and leading to data breaches or incorrect outcomes.
4.
Over-reliance and Loss of Human Oversight
-
"All-Green"
Fraud Scams: AI systems can fail to detect fraud when scammers manipulate
customers into voluntarily moving funds, as the transaction looks
"normal" to the AI.
-
Errors
in Judgment: Relying entirely on automated tools without human oversight
(or a "human-in-the-loop" approach) can lead to mistakes in approving
transactions or client assessments.
5. Regulatory
and Compliance Challenges
-
Evolving
Regulations: As AI technologies advance, regulatory frameworks struggle to
keep pace, creating ambiguity and compliance risks.
-
Legal
Consequences: Failure to comply with regulations, such as the EU AI Act or
local data protection laws, can result in severe fines, reputational damage,
and operational bans.
6. Data
Quality and "Hallucinations"
-
"Garbage
In, Garbage Out": Inconsistent, fragmented, or low-quality data leads
to inaccurate AI insights and erroneous financial predictions.
-
Hallucinations: Generative
AI tools can create fabricated information or incorrect, yet confident,
financial insights, causing significant risks when used in investment
decisions.
To address
these pitfalls, institutions can prioritise several risk mitigating measures,
to name a few think of:
-
Implementing
models that allow for clear audit trails.
-
Actively
remove bias from training data and use diverse datasets.
-
Requiring
human review for decisions like loan approvals and large, non-standard
payments.
-
Establishing
internal AI ethics boards and strict data security protocols.
And of course…
train all staff and ensure awareness is taken seriously.
