Research Paper: AI, the birth of a new path

By: Taran Krishna Thummalacharla

AI is a marvellous tool that has taken the world by storm, with a variety of uses and can act as a
compass which can direct individuals into a larger database filled with knowledge and skills
beyond previous limitations. Hence we attempt to explore the effects of AI within the finance
sector, a sector that has seen these impacts firsthand and has been known to see different displays
and uses of AI for the sole purpose of the improvement of the financial sector and personal
finance as a whole, the limitations and challenges that come with as well as the overall benefits
that AI is able to influence within the realm of finance.
First we must highlight that AI is one that has been used by multiple companies within the
finance sector for a varied uses and for a variety of reasons, take for example the global financial
symbols of PayPal and JPMorgan implementing AI and machine learning into developing a fraud
detection system or Betterment, another financial sector symbol that is using AI in order to help
reduce taxes within transactions as well. Within a small span of time, multiple symbols within
the finance sector have been seen to adopt ideas of using AI for their growth, with CEO of
PayPal Alex Chriss even going so far out as to mention that they plan to spend 2024 as a
transition year to build up using AI.
Now the clear aims that this research paper plans to bring onto the table is one full of raw
statistics, ideas, changes etc. on the varied usage of AI in the finance sector

Key Ideas:
1) Applications of AI

 Fraud detection and cybersecurity

 Market predictions and algorithmic trading
 Risk evaluations
 Personal finance tools

2) Real world instances

 PayPal on fraud detection

 Wealthfront and Betterment for personal finance
 BlackRock’s algorithmic trading platforms

3) Challenges and ethical considerations

 Possibilities of Bias in AI programmes

  Data security and privacy concerns
 Job displacement and economic instability
This paper plans to encapsulate all these ideas using previously made case studies, surveys, and
infographics as well as viable news articles and sites in order to portray these ideas.

Historical context:
Before diving into the key ideas for this research paper, the historical context of AI within the
finance sector will first be explored in order to begin our journey into the impact of AI within the
finance sector.
Despite it trending impeccably in the modern era, the term AI was first coined within 1956
within the Dartmouth summer research facility, while the surge for competition between AI
amongst companies had manifested all the way within the 1980s, where more than 66.6% of
fortune 1000 companies had begun their development within AI.
Yet in 1982, that’s when AI’s presence was first truly felt throughout the finance sector of the
world, when James Simons founded a hedge fund, Renaissance technologies. As of November
2024, this has a market cap of 14.86 billion$, with reverberating effects echoing across the
finance sector from that point forward. Many more project like so take place 4 years later such as
Dupont investing in 100 expert systems, saving him 100 million$ that year or APEX
becoming the first system for financial plans.
In the 1990s, the mitigation of fraud detection had suddenly become a fad of sorts, here enters a
new financial AI system, the FinCen Artificial Intelligence System or FAIS for short, this system
was able to review over 200,000 transactions a week. Being able to check over 400 potential
money laundering cases over 1 billion$.
Entering the 21st century, huge developments within AI for the realm of finance were made, such
as in 2003 where rule-based system for financial intuitions for corporate loans were made
by Hodgkinson and Walker or in 2013 where Janulevicius and Goranin developed a similar
system made specially for small and medium sized companies for data research on security
threats. Even though expert systems don’t exist within the realm of finance anymore, these are
considered as an ignition for AI within finance and the modern world.
AI has become more impactful near the end of the 2010s and the beginning of the 2020s, where
a study had been conducted which implies that as of 2020, there were around 52% of
financial services having AI implemented processes, within each one to provide quick outputs
towards any financial advice that had been requested. In 2023 as well, AI had superseded itself
within FinTech, with their AI’s market size growing to 11.89 billion$ with the industry’s
overall AI spending being at 35 billion $.
The future of AI continues to carry heavy expectations with it as Accenture predicts AI
augmented technologies to aid in financial services with a value of over 140 billion$ of output,
within the financial sector by the year of 2025. Even traditional financial institutions are to soon
be shown to have an overall 1 trillion$ cost reduction from AI in 2030. AI’s value has also
been predicted to enhance from billions to trillions within a decade as Accenture also reports
that AI is set to boost the financial industry to 1.2 trillion$ in the year 2035.
AI in finance had been adopted for 3 obvious reasons, their endless abundance of data, their
ability to comprehend different problems and to make improvements at rapid speeds.

Fraud detection and cybersecurity

This research paper commences with the first key topic being about AI’s ability to enhance fraud
detection and cybersecurity, some examples of AI being used in these ways can be seen with
PayPal implementing machine learning into their system in order to improve against potential
money laundering attempts from scandals that utilize AI in order to participate in these events.
Fraud is considered as a heavy challenge against finance as a whole with 1.08 billion dollars
being stolen within the year 2023 itself, proving to be a tedious task for AI to help mitigate.
By doing this, fraud must be understood by the distinct types they come under and what exactly
they do, these three types are sign up fraud, login fraud and payment fraud.
AI’s current abilities and progress within fraud detection involve its use of machine learning in
order to make timely analysis of the situation where fraud has occurred for a particular user, first
through the improvement and increased implementation of AI within fraud detection.
AI is able to make an impact within fraud detection due to the real time transaction monitoring,
AI’s behavioural analytics towards fraud detection and detection of payment fraud through
copious amounts of databases. AI is also evolving towards a state where it is able to track
consumer spending patterns in order to understand whether fraud is occurring or not in
order to act. Increased use of AI driven financial products has also installed these features in
order to mitigate fraud as well, AI also looks forward to seeing a surge of AI involved platforms
which consider automated wealth management in order to avoid incidences of fraud as
well.
Machine learning technology is hence given immense importance when considering the
implications it brings onto this task, this is starting to prove itself as 18% of pros count the use
of machine learning in anti-fraud fights in their fraud detection tools where a separate 32%
plan to use this technology within the span of the next 2 years. These stats show the use of
these programs being nearly tripled for fraud detection by the end of the year.
Biometrics and robots have slowly seen a steady rise in their demand for fraud detection from
2019 as the use of physical biometrics had increased to 14% within 2019. Uses of these
technologies are often seen within banking and other similar financial services, where 51% uses
physical biometrics whereas another 33% use robots.
AI specializes in the following when considering its effectiveness in fraud detection
  Transaction analysis: Where AI is used to analyse the spending methods of consumer
behaviour relatively quickly in order to find suspicious behaviour on an account, a study
by SAS shows AI being able to find 500 transactions per second in order to
implement transaction analysis which later follows into saving significant losses
 Authentication and verification: AI verification processes often undergo instances of
reducing payment fraud through ensuring tha the transaction is being done by the account
that it is supposed to be done by the account owner. AI is able to process around 95%
of biometric verifications within less than 10 seconds
 Data collection: A function of AI used widely, beyond finance as well but in this
instance, is used in order to collect multiple sources of behavioural data within
transactions and apply it to its other functions
 Machine learning: Algorithms are trained in order to help recognize changes in spending
patterns, supervised learning datasets are also used to check the difference between
legitimate and fraudulent transactions with increased training by tracking historical data
from the account. Machine learning can do this by putting a risk meter for fraud detection
as well, with a risk score on a scale of 1-100, where the higher the score, the higher
the risk.
The use of AI of fraud detection often lies within different cases that relate heavily to finance,
being able to diversify itself as well, fraud detection of AI thrives most in situations such as
banking and financial services, within e-commerce for customers as well and within virtual
economies. This is due to one common factor that resides amongst all three cases which is that
they all require a digital presence, making it easier for the AI to be able to track it as well.
Benefits of AI:
AI in fraud detection also brings a numerous amount of benefits such as its improved efficiency
in being able to track down fraud and the accounts that have been affected by it as well as the
scale at which it continues to be applicable to as the more transactions that are made, the more
the scale that the AI measures it upon grows with it as well.
AI also has the advantage of being able to last 24/7 and track transactions in real time, hence
making detection much more efficient and letting the response towards fraud be executed at a
faster rate.
AI’s capacity to analyse different data also lets there be more accuracy within its responses
and reporting it more easily in order to take action against fraud, AI is more likely to make less
mistakes due to its wide range of information in comparison to us humans, hence making it much
more efficient to track down the different spending patterns and transactions more easily than a
human could.
The main benefit with AI being used in order to aid against fraud through fraud detection is its
ability to adapt using machine learning, as with each transaction, AI is able to apply new
tactics and ideas onto identifying fraud and how it can be mitigated as well, adapting to patterns
by scammers.
Challenges:
The efficiency and ability for fraud detection to be worked upon depends on the ability of the AI
to collect and store data, if it is unable to then data can remain biased or incomplete when going
over fraud detection. The quality of data within AI can also prove to be insufficient in order
to track fraud if the spending patterns are similar which is what machine learning is often
dependent on. Privacy concerns and different regulations regarding how available data can
be also arises, causing different ethical issues when using AI to track spending patterns.
Fraud tactics also improve within the time it takes for an AI to properly analyse the
patterns that occur with fraud, fraud detection hence becomes harder with machine learning as
it will be forced to adapt to different fraud patterns but not quick enough before fraudsters can
change their patterns to other ones in order to avoid fraud detection.
Integrating AI into different fraud detection systems where legacy systems may not be
compatible with AI and its machine learning, the integration process in and of itself utilizes a
vast range of resources as well while the functionality of the system can be downgraded
during this transition period, resulting in fraud becoming more present without being able to
track it as efficiently as before.
Another major drawback about AI is that it is common within being used for fraud as well,
being used as deepfakes in order to take over different accounts that require biometrics in
order to sign in, AI can also grow in order to combat against AI being used for fraud detection
considering that it has been implemented earlier, giving it more time to analyse different possible
patterns for fraud detection. Deepfakes regarding AI has shown to have an increase of 2137%
in fraud attempts across the past 3 years as well with the implementation of AI in fraud.

Market predictions and algorithmic trading

The stock market is all the rage when people start considering what to do with their money and
income, or rather when discussing the various income sources possible for them to generate
money the fastest way of which they can think. Stock markets also present themselves as a
source of money that can be very versatile or a liquid form of money, acting as a current
asset for people and even shares of different companies being able to act as collateral depending
on their value.
A combination of two rising giants would hence make a large ripple in the realm of finance as a
whole with the effectiveness of AI and the versatility and liquidity of the stock market, yet what
are some of the main components that AI plays within influencing stock market predictions and
making algorithmic trading more possible for people to learn from as well as for AI to allow
people to make a profit out of these predictions and trading. Another thing to mainly discuss is
the progress and impact that AI has made onto market predictions and algorithmic trading, hence
influencing the direction at which this section of the research paper aims to explore.
Features
AI as of current is able to make long lasting impacts onto the market and through algorithmic
trading through features such as the ability to operate in real time and analyse data within
that time span. Normal algorithmic trading systems are capable of processing data within the
stock market across regulated boundaries and directions. AI enhanced systems on the other
hand have the ability to work with encapsulating new data by the minute, trading with AI
also allows it to be used as a type of monitoring tool as well in order to view and predict changes
in the stock market. AI uses natural language processing in order to determine the best decisions
and enable full efficiency in market predictions.
AI also makes trade more efficient by eliminating possibilities of human error within the
decision-making process, AI can do this by using natural language processing or NLP for
short by eliminating unnecessary information and segregating the useful information from
it in order to make these decisions. By doing this, AI is able to minimize the effect of human
bias and error when entering the stock market.
Algorithmic trading with AI is boosted due to AI powered systems being more efficient by
executing more trades, analysing vast amounts of data rapidly for the traders to make a
more informed analysis, mitigating possible risks in portfolios and evolving in different
environments as the stock market fluctuates, by evolving into different environments.
Algorithmic trading is able to use these features in order to make more efficient and informed
decisions for traders to execute.
Scalability in AI is one last feature that truly makes AI within the stock market lethal, using
traditional trading as an example, the number of trades that traders can do through monitoring
each trade is relatively time taking and inefficient but with AI, traders can extend to numerous
markets at a faster rate with a large amount of data when entering those markets, making
the level of activity within the market much more efficient and with more execution.
Statistics regarding the impact of AI on the stock market
The AI global market is often the first thought that comes to mind when discussing the impact of
AI on the stock market as a whole, yet in this section this paper also aims to tackle how
algorithmic trading has developed a progression into the financial sector through various
different case studies and statistics in order to determine the effect of AI in such a way.
Now the global AI market for reference is predicted to be at 1.81 trillion$ by 2030, becoming a
major factor concerning the increase in AI within algorithmic trading platforms, which started to
rise in 2017. Increase in AI within these platforms spiked in 2019 with a 19% increase in AI
elements within algorithmic trading platforms and then with a wave striking, showing over
50% increase in these elements from the 2020s, marking the development of AI in algorithmic
trading programs. Though they remain smaller compared to other ETFs (exchange traded funds),
AI has seen a surge in the increase of their ETFs within March 2020 to be allowed, AI driven
ETFs also operate within once a month, contrary to normal ETFs that do so in a year, creating a
more liquid market as well.
Institutions with high standing in the financial sector also see reason as to how AI can make an
impact into algorithmic trading such as JPMorgan leveraging the use of AI for market
segment analysis, portfolio management and optimization as well as risk management.
Goldman Sachs also acts as an investor into the potential AI can bring into algorithmic trading
and market predictions by employing different AI algorithms to streamline in a high frequency
for quality trading, these systems were deployed to help Goldman Sachs to minimalize their
losses when in volatile periods using risk management. Renaissance technologies had also
adopted ideas of using AI within the stock market by relying on its Medallion fund, a fund run by
AI in order to analyse complex data and process it at high speeds, by using machine learning, the
fund now identifies different patterns within the financial markets to contribute to a more
sustained ROI (return on investment) by reducing human error.
Challenges of using AI in algorithmic trading
Certain challenges that arise when inputting the capabilities of AI into algorithmic trading
systems involve the following
 Risks of investing and overlifting: This occurs due to losing money potentially due to
investments within securities and other financial products and a loss of generality which
can contribute to a mediocre performance and a decrease in decision making due to being
unable to respond to rapid fluctuations in the stock market
 Hacking: Algorithmic trading programs that use AI are susceptible to hacking,
manipulating these programs to cause for unprofitable trades and losses for investors and
traders alike
 The Black Box: This process involves the opening of complex algorithms, making it
difficult for traders to comprehend the reasons behind a particular decision, a lack of
transparency with AI causes traders to trust it less, by not being able to understand it,
traders won’t be able to trust AI.
 Insufficient or the quality of data is lacking: When AI encounters inferior quality of
data, this can affect the decision-making process for an AI within the stock market,
causing it to make bad decisions that ultimately end up in a loss.

Risk evaluations
Even in previous sections, a key factor about the ability of AI to prove its efficiency within
aspects of finance often involve AI being able to thrive within analysing the different
risks that may occur within different situations and plans as well.
Risk evaluation often takes up personas that involve overseeing details that the normal
human wouldn’t be able to do when using AI, such as minor trends and shifts in patterns
within the stock market. AI is able to achieve this by using machine learning models yet
again to assess and enhance the accuracy of predicting risks, such as those with credit,
and the ability of financial institutions to assess credit.
 Studies have even displayed a 40% increase in organizations within increasing their
investment in AI in order to continue with advances in gen AI, by employing different AI
for risk evaluation, they are able to showcase their ability to analyse the financial market and
predict market volatility much more efficiently compared to that by humans, with a large and
vast database, AI can easily respond to any changes within the market and develop new
patterns for it. Services regarding AI hold around 39.64% of AI market share within
2022, increased investment into AI makes it much more of a necessity in order to
increase efficiency in risk evaluation.
Tools
Some tools regarding AI’s ability for risk evaluation involve the following
 Anti money laundering compliance (AML): This AI application is used in order to
scrutinize and detect any suspicious transaction patterns, AML is used in order to
improve the speed and detection towards assessing client databases globally. By
using AML, banks improve their ability to check for suspicious activity by 40%.

 Risk management in insurance: AI uses its large database in order to assess the risks
that a client portfolio may have and assesses the possible write-offs and decisions it
might be able to give to customers and clients. AI is able to assess risks more easily due
to this process. AI is now able to reduce the time to assess risk within insurance by
90% and is reported to have a 25% increase in the accuracy of their predictions. AI
underwriting techniques also decrease costs by 50%

 Portfolio risk optimization: A clear feature that AI is able to use in order to control the
risk management within portfolios is through asset allocations by judging different
trends, economic indicators, and the company’s performance. AI can hence shift from
safer to more profitable yet risky assets by analysing these market trends. 91% of
managers already consider (54%) or plan to (37%) utilize AI within their asset
plans. A study in 2023 finds that a use in AI for portfolio management reduces
portfolio illiquidity by 10%.

Personal finance tools

Finance, the first thing that arises to ones’ mind with this word is their capability and capacity to
earn and store money, whether it be for a company or for a local man on the street, finance
always boils down to the ability of an individual to maintain their own personal finance. With the
growing prospect of AI becoming more inclined within completely managing financial burdens
that take too long to check manually.
When considering the ability of AI to utilize and manage personal finances, talking about
personal finance brings up two key topics to discuss about, regardless of AI involvement or not
which are the ability for an individual or individuals within budgeting and the same ability
within investments. After analysing these key factors, the ability for AI to execute these
requirements to the fullest, truly display the ability for AI to act as efficient personal finance
tools. So far, it seems that the adoption of AI within financial functions have increased from
the past year with 58% using technology from a previous 21%.
Budgeting
AI tools that involve budgeting are of different varieties such as automated budgeting,
predictive analysis, rocket money etc. The three tools that will be explored when considering
the ability and efficiency of AI within personal finance and within budgeting will be automated
budgeting, personalized budgeting, and credit monitoring tools to tackle various aspects of
budgeting and determining the ability of AI to satisfy such requirements.
 Automated budgeting: Automated budgeting with AI is the ability to make personalized
spending and saving patterns and tailor them to the income that is generated, AI is able to
do this by analysing different consumer patterns and linking it with their income which is
done through machine learning algorithms, used to inform customers about their
spending and saving habits to help give a complete overview about their financial
situation.

 Personalized budgeting: AI can enhance the personalized budgeting of an individual by

controlling the data from the individual’s spending patterns and their income streams, this
form of budgeting allows AI to make more empowered decisions after taking these
factors into account. Using AI for personalized budgeting allows an individual to
make a more informed decision rather than that of a more proactive one. This tool
also allows AI to make different savings plans and decisions for the individual based on
their income and help make informed decisions about the amount of income that is to be
saved. This tool has been predicted to accumulate 2.16 billion$ in 2028 at the current
annual compound rate (9.99%)

 Credit monitoring tools: Like many other financial applications of AI, credit monitoring
tools install machine learning in order to evaluate an applicant’s database against the
other patterns that the AI has learnt over training, this occurs when a credit application is
received by the AI system. This helps to generate a score that can make the customer
notified about their credit and their credit worthiness, giving customers innovative
ideas about how credit scoring may work for them as well.

Investments
During managing the income that is generated by an individual, ideas that often come to mind
are the amount that they need to spend, save, and invest and the need for that individual to spend,
save and invest. After covering the ability of AI to manage the spending and saving of income by
an individual, specific tools that can cover the ability and the amount that can judge the ability of
an individual to invest will now be looked upon. Like with budgeting, a lot of common tools
arise when considering AI tools that are capable for investing, yet the three that will be used in
order to determine the efficiency of AI within personal investments will be through sentiment
analysis, factor investing and debt management.
 Sentiment analysis: This refers to the ability of an AI to evaluate and identify large
trades and high-speed trades within the market using natural language processing
(NLP), based on the personality of the trader, where the AI is able to develop new
patterns and stocks that the trader can invest in based on the type of trader that
they are. By evaluating the sentiments that traders go with into the stock market, an AI
can help mitigate the human error possible when investing in certain stocks by making
more informed decisions.

 Factor investing: This refers to the quantitative and quantifiable investment approaches,
AI can do this through assimilating different risks whilst product complexity
continues to rise, with this AI can hence make different counteractive trades in order to
attack different market scenarios to take care of losses and gain advantages within the
market. Assorted products and strategies are tied to the strategic analysis that an
LLM (large language model) can decipher.

 Debt management: The application of AI across personal finance extends towards debt
management and credit optimization through analysing user financial transactions, their
debts and their spending habits that are used in order to reduce debt. AI repayment
using debt depends on strategies by AI such as personalized payment
recommendations to tackle with the most effective payment strategies, debt
consolidation options in order to transfer loans and balance credit card transfers to
mitigate extra interest payments. By prioritizing debts with high interest rates, this tool
then suggests the different payment strategies after identifying these debts, to save on
users’ interest money as well.

Real world instances

This portion of the paper will aim incorporate areas where financial tools discussed previously
are being used as of current in order to make AI improve the financial sector in the most efficient
way it can, the ability of different financial institutions to implement these tools for the benefit of
global finance and their customers as well as address how the market share of AI increase as
these real world examples that utilize these financial tools are taken into account by different
financial institutions.

PayPal on fraud detection

PayPal is a large financial institution which rules over the financial sector, acting a s a key
symbol on what it means to be able to display the ability to control and utilize finance, when
considering the applications that can be done with AI, a very popular case study within the
financial sector about PayPal in fraud detection arises with PayPal’s ability to enhance the
capabilities of AI within fraud detection by analysing the following and how AI has increased
fraud detection against AI and with AI.
 Believing that common protection methods may not suffice, PayPal has made the idea of
implementing robust authentication centre against any fraud detection that has been
using AI.
 PayPal has also implemented a new type of algorithm, known as PayPal’s Fraud
Protection Advanced, in order to help merchants with their fraud analysis to protect their
business
 PayPal has shown their belief in the efficiency of AI for risk management by teaming up
with the StepStone group to invest $30 million in a generative AI start-up company
known as Rasa to help control the level of risk management
 PayPal is now using machine learning, a subset of AI, that uses algorithms to scan for
large amounts of data for different spending patterns and insights in them, to then apply
into making better decisions for the programmer towards better perception and decision-
making ability of the computer system, in the context of e-commerce, machine
learning is a type of AI that is used for payment detection fraud.
 Some types of AI that have proved especially help against fraud, these are supervised
learning, unsupervised learning, semi supervised learning, and reinforcement
learning.
 Supervised learning describes AI which are programmed for the purpose of running
through predictive analysis with previous data that is labelled as good or bad.
 Unsupervised learning uses an AI which analyse untagged data and identify any
anomalies within customer spending patterns
 Semi supervised learning is using models to split the differences between approaches of
supervised and unsupervised machine learning, using fraud detection algorithms to carry
out this process, removes human error from the process as well.
 Reinforcement learning allows the AI to learn all optimal behaviours within the
environment around from various interactions to then detect such fraudulent anomalies.
 Examples of finding such fraud with machine learning would be sign up fraud: when a
scammer creates a new bank account with a stolen identity, login fraud: taking over an
existing account, payment fraud: when scammers utilize a credit card without the
owner’s knowledge.
 Machine learning is able to aid this way by analysing third party information, monitor
devices, IP, phone transaction and any data which may suggest any anomalies or fraud,
machine learning is also able to rapidly assess whether the individual is a verified user
PayPal implements machine learning mainly into their usage for determining fraud detection at a
more efficient rate, by doing so, PayPal can also ensure that their customers are able to make
transactions with a higher sense of security in comparison to other financial services from other
large financial institutions resulting in PayPal gaining more customers by implementing these
factors.

Wealthfront and Betterment on personal finance

Wealthfront is an investment services company which is situated in Palo Alto, California and
was first founded by Andy Rachleff and Dan Caroll in 2008. Betterment is a separate
American advisory firm that provides digital investment and cash management services, this
firm was based in New York and is registered with the securities and exchange commission.
These companies specialize within financial services regarding personal finance of their
customers and providing these services tailored with expert efficiency in order to make the
necessary decisions required for their customers through the increased use of AI in order to avoid
human error. These are achieved by the following features and uses of AI within aiding their
customers with efficient ideas regarding personal finance.
 Hyper-personalization: Algorithms where AI can analyse a large amount of data about
the consumer’s financial situation and the risk that they are put at, Wealthfront has
obtained a robot adviser, made to create personalized investment portfolios for
clients who require to track their risk tolerance and their financial situation.
Betterment is also able to achieve the same by using similar robot advisors in order to
offer different investment strategies to their customers, their AI algorithms are made in
order to monitor portfolios and balance investments by overseeing the market and
individual goals.

 Behavioural coaching: Using AI in an analogous way to that of sentiment analysis

where the AI is made to provide personalized nudges towards investors and
individuals towards the decisions that they will be guided towards more emotional
yet safer decisions for investors.

 Methodology: AI is used in order to manage different portfolios by managing the clients’

assets, their investment vehicles etc. An AI is able to achieve to this by using a variety of
factors as mentioned previously in order to determine the priorities within different
portfolios and what is needed most within each portfolio, using the different sections to
make a methodology to follow in order to execute this. Wealthfront does this by
incorporating a Black Litterman model which is for assets which a model made in
order to match their risk tolerance with the expected outcomes. Betterment also
achieves comparable results using a similar model in order to track these assets, tailoring
mainly in personalized investment strategies.

 Personalized investment strategies: This is where an AI drafts a portfolio and strategies

for investors towards personalized investment management, Betterment is able to
 achieve this by incorporating their AI system with tailored investment strategies
regarding possibilities of loss and rebalancing. Wealthfront is able to achieve
comparable results using robot advisors in order to draft portfolios in such a way.

Wealthfront is a company that incorporates AI into its system for personal finance
management through the use of robot advisors to oversee the primary concerns within a
portfolio and how to mitigate them as well as use the AI in order to develop different algorithms
and systems that are tailored to match the priorities for a customer within the financial portfolio.
This involves their use of the Black Litterman model and their use of algorithms in order to
provide advice about rebalancing investment and developing different strategies as such.
Betterment is a company that incorporates AI into its system for personal finance
management by developing AI systems which are tailored towards sentiment analysis and to
create personalized financial strategies and nudge investors towards stocks and shares that
benefit their portfolio best while taking into consideration of the market situation and the type of
investor.

Blackrock’s algorithmic trading platforms

BlackRock is a global provider of investment services, advisory services and risk
management, BlackRock acts as an asset management in order to provide insights to their
customers about the different risks and benefits that come with investing in a particular stock,
BlackRock mainly specializes in investments, hence developing an AI algorithmic trading
platform for this reason.
With a rise in large language models (LLM), BlackRock uses these language models for the
purpose of creating different chatbots in order to be used for security analysis and trained
within different datasets for investment tasks at a high degree. BlackRock also utilizes a thematic
robot tool in order to blend in with human insight and using LLM’s to build up a larger
database with equity baskets that are seen with a greater efficiency. LLM’s and the thematic
robot model will be crucial in going over BlackRock’s algorithmic trading platforms and
determining how AI is used in it.
 Large language models (LLM): These play a key role within the algorithmic trading
platform made by BlackRock by efficiently analysing and synthesizing information from
various sources, LLMs are trained to process large elements when accounting for
different contextual relationships. LLMs in the algorithmic trading platform made by
BlackRock are used to contextualize the potential for more accurate differentiated
investments. LLMs are often integrated this way through chatbots, these are used in
order to tune the investment processes and perform different investment tasks based
on the market forecast and how the LLMs are able to interpret it. LLMs are often
trained for more specific datasets in order to carry out these tasks with prominent levels
of accuracy.
  Thematic Robot: By integrating LLMs into different equity baskets, a “robot” is used
to blend in the ability of LLMs towards data that is able to build up longer or
shorter equity baskets. This is done as investors are often in need of building up
security baskets which can be time-consuming for broker baskets that aren’t built well
enough. These thematic insights are often used in order to build up a portfolio, these
insights often involve using an LLM in order to measure the risk of current
portfolios to more emerging themes and to give an insight towards companies that are
directly linked with this thematic basket as the main objective for this tool is to be able
to bring speed and accuracy to the construction of the thematic basket.

 Accuracy: BlackRock’s main objective within their AI algorithmic platform is to be

able to utilize LLMs in order to bring out accurate insights towards earnings and
portfolio development. BlackRock has proved to do this with their proprietary model
that is made to be trained on 400,000 different earning transcripts that span over
17,000 public firms. BlackRock has also compared their model’s accuracy to that of
OpenAI’s GPT models, with the BlackRock systematic coming out on top with more
than 0.6 accuracy in comparison to GPT’s 0.55.

BlackRock’s algorithmic system overall looks at the ability of their system to analyse
quantitative information after collecting data from nearly 2 decades, this is used in order to create
more precise yet differentiated investment outcomes and ideas. BlackRock aims to do this
mainly with the usage of LLMs to analyse substantial amounts of data and to build
thematic robots for different equity baskets in order to look for better investment outcomes
or ideas with a key highlight on their LLM’s accuracy.

Challenges and Ethical considerations

Often with every gift and miracle that is bestowed upon the world, so comes an equivalent bane
to accompany it, throughout this research paper, the consideration of the ability of AI to aid the
financial world within a versatile range had been considered yet with those come challenges that
present great distress at a global scale, some that can be mitigated, some that cannot.
The ability to perceive AI as the ability to utilize humanity’s creations to its fullest potential is
what truly determines how one can overcome these arduous challenges, job displacements are
often discussed when considering the repercussions that occur with using AI. Others that come to
mind is the use of AI against personal data and on problems that may occur such as the Black
Box problem. Others consider the use of AI in fraud and hacking.
All of these problems are to be explored further down this segment with key ideas regarding the
challenges that AI presents within the realm of finance as well after going over the benefits that
AI brings withing finance as a whole.
Key Ideas
 Discovering any ideas towards possibilities of bias in AI programmes, possibly due to the
developers of these AI
 Data security and privacy concerns regarding AI, with a vast database, it becomes
concerning on how much AI knows and what ethical concerns it raises
 Job displacement and economic instability, a huge determinant on the utility of AI when
moving forward with it in the financial sector

Possibilities of bias in AI programmes

Bias refers to the systematic errors and inaccurate decisions regarding a personal feeling that is
what makes algorithms and models be considered as biased, these biases can be done within
development and training, often leading to outcomes which may not be too favourable,
impacting the decision-making process that comes within correcting the overall integrity of
different financial systems.
Distinct types of bias can arise within Ais amongst the development of the AI or within the data
that the AI interprets and bases its views upon, those types of bias arise especially when
incorporating a vast range of information into the algorithms. These types of bias that can
occur within AI are data bias, algorithmic bias and human bias, these biases incorporate
mainly problems that arise within the developing bias, with special emphasis on human bias as a
whole.
 Data bias: When the training model of the AI with the training data to build up the AI
model consists of bias that could be present from the developer or from the source of
information used. An AI then amplifies these biases which can lead to more inaccurate
and depreciating errors in finance. USC researchers recently found that bias is present
in 38.6% of ‘AI facts.’

 Algorithmic bias: Different algorithms and models that incorporate AI can lead to heavy
biases if not treated carefully at the developing stage of the AI model, AI algorithms like
these rely on data as well, algorithms will be influenced by information that occurs
within a particular set of data and makes their judgement about another set of data in the
same way. Out of 555 AI models, it is found that 83.1% of them contain a high
proportion for risk to bias.

 Human bias: This type of bias is very self-explanatory, it talks about the human bias
that occurs during the development stage of the AI model even when considering the
data and algorithms the AI containing being free from bias themselves.

Data security and privacy concerns

AI has multiple different concerns regarding privacy and data security overall, this can occur
through different financial mediums such as banking and misusing financial data of a
consumer of a financial service that is run by an AI.
Distinct kinds of privacy concerns and risks that are to be considered when using an AI for
financial services and mediums are data misuse, data discrimination and data loss, these
involve AI’s making biased outcomes or using the data stored for purposes that aren’t
tailored towards the desired outcome that is needed by the owner of that data. These occur
often in financial services regarding accounting and finance as a whole.
Potential financial fraud with AI also raises different risks and questions on AI’s misinformation
within the market and AI not being able to collaborate with regulations and legal boundaries that
the AI is required to act within, this kind of behaviour often challenges how AI is able to
harm the financial stability of an individual with the potential of committing financial
fraud.
Concerns regarding AI’s ability to function within the finance sector and its system is also a
heavy determining factor on the types of danger that AI is susceptible to and what it may
cause with the data at its hands. This involves AI’s possibilities of algorithmic bias,
vulnerability to data quality issues and accuracy of AI predictions when considering AI’s
capabilities under financial advisory systems such as Betterment and Wealthfront. Financial
institutions tha use AI in similar ways to these firms can be victim to faulty decisions that are
made by the AI and lead to a loss economically, AI’s complexity as a whole is considered to
be a factor for these challenges and risks as well.
The overall loss of the average short term in abnormal returns made by AI’s in banks and
other such financial services is around -21.04% and the overall negative repercussions that
follow it through the financial industry is -0.13% as well. This data was collected in the year
2024, stating as well that financial institutions such as banks and other such financial services are
more suspect to AI incidents with high bankruptcy and operating at lower cash flows. A similar
internal AI incident within calculation errors for mortgages under a modification and
underwriting tool were made, this incident resulted in 500 people lose their homes and
increase the number of loans required to mitigate it by 100s in 2023.
AI incidents have led to losses within profitability, the ability of the AI to capitalize on the
market, within calculation errors for loans and mortgages and overall proving to be a
threat through the privacy concerns that it creates with its large database as well as other
debates on whether AI must be used for financial services such as banking after such incidents,
considering the vulnerability of AI as learning models or through the bias that is input within
AI during its developing process that will steer individuals into more risky or less secure
decisions regarding their personal finance. After such incidents, questions about AI’s ability to
store and manage such a large amount of data is also to be discussed upon with such frequent and
significant errors being made.
Job displacement and economic instability
AI’s main concern that it brings when introducing its capabilities in the finance sector and the job
market as a whole involves AI replacing jobs that firms may find to be trivial or rather
unnecessary for them to maintain wages for employees that have jobs that can easily be filled by
the abilities that AI is able to do at a diverse range with a higher level of efficiency. Economic
instabilities follow suite as unemployment rises within these sectors, resulting in an uneconomic
use of resources through the form of labour.
Through historic rates regarding labour shedding, 1 to 3 million jobs could be potentially lost to
AI, regardless of the significance of these jobs, though these displacements will occur gradually
they will increasingly grow as AI adoption amongst different sectors and companies increases.
On an annual basis, this results in job displacements occurring at a rate of 60,000 to 275,000 job
displacements a year.
When discussing about AI’s effects on economic stability, the thought of AI’s ability to affect
unemployment and the labour force is what comes in mind, with AI being more frequently used
in firms, this is seen as within the next 20 years, around 54% of the jobs within the European
Union are at risk of being victim to job displacement. This involves routine jobs that can easily
be displaced by AI within larger companies that store a huge database of information and
resources.
AI also affects economic stability through potentially influencing wages, the income distribution
amongst workers and increases in economic inequality, the international monetary fund has also
made statements and predictions on AI’s ability on jobs, stating that at least 40% of jobs
throughout the world will be affected by AI. AI has also been seen to complement higher income
workers, leading to increased inequalities in income as AI is more so forecasted to displace jobs
within the lower wage class, such as jobs that are considered as routine jobs, though resulting in
a gain through the form of productivity, economic inequalities are increased with the introduction
of AI towards different job displacements, causing increased redundancy amongst companies.
Economic instability and job displacement are severely affected through AI’s displacement of
jobs that are considered more routine or rather less skilled, AI through this way causes for wages
to not be properly determined and increasing inequalities by taking more lower income jobs.
Jobs around the world within the lower paid is seen to have around a 55% exposure towards
these jobs within more advanced economies as AI is more likely to replace these routine jobs as a
form of factor substitution for the sake of productivity, yet this in turn leads to more resources
becoming an uneconomic use of resources due to increased unemployment as advances in
technology increase.

Conclusion
Finance as a whole incorporates the ability of an individual or institution to manage, utilize and
continue with the circulation of money through assets or through cash flow, the ability of AI to
accomplish these tasks on a numerous and versatile scale with increased efficiency in
comparison to completing these tasks that are highly susceptible to human error. AI has been able
to demonstrate its abilities in finance by being able to act for financial institutions and for
personal finance as well by incorporating its large and vast database into multiple algorithms.
Examples of AI being used frequently for these purposes involve PayPal’s integration of AI
within fraud detection for improving their services for their consumers, and the use of AI within
investment algorithms by Betterment. AI has been shown to have a diverse range as such by
stretching across the security within finance towards the applications of certain decisions based
on a large base of data where decisions are made through market patterns.
Even though AI has come with numerous amount of challenges that range the same way, AI’s
benefits through its growth with machine learning and large language models can help to
mitigate and correct these negative externalities that AI causes when introduced into the finance
sector.

Citation
Introduction:
 https://thesocialshepherd.com/blog/ai-statistics
Historical background:
https://analyzingalpha.com/history-of-ai-in-finance
https://eimf.eu/how-will-the-influence-of-ai-affect-the-financial-sector-in-2024/
Fraud detection and cybersecurity:
https://www.paypal.com/us/brc/article/payment-fraud-detection-machine-
learning
https://www.researchgate.net/publication/
384931095_Artificial_Intelligence_in_Fraud_Detection_and_Financial_Risk_
Mitigation_Future_Directions_and_Business_Applications#:~:text=The
%20data%20indicates%20a%20significant,positives%20and%20increasing
%20detection%20accuracy.
https://fingerprint.com/blog/ai-fraud-detection/#:~:text=Identity
%20verification,builds%20trust%20for%20your%20users.
https://www.digitalocean.com/resources/articles/ai-fraud-detection
https://www.evertecinc.com/en/the-role-of-artificial-intelligence-ai-in-fraud-
detection-key-statistics-and-applications/
https://www.sanctionscanner.com/blog/how-does-ai-powered-id-verification-
fight-digital-fraud-711#:~:text=Businesses%20can%20also%20use
%20video,produce%20real%2Dtime%20verification%20results.
https://www.ravelin.com/insights/machine-learning-for-fraud-
detection#:~:text=Using%20machine%20learning%20to%20generate%20a
%20fraud%20risk%20score,-Customer%20places%20order&text=At
%20the%20point%20of%20the,higher%20the%20probability%20of
%20fraud.
https://emerj.com/artificial-intelligence-at-paypal/
Market predictions and algorithmic trading:
https://www.utradealgos.com/blog/top-6-ways-ai-enhances-speed-and-
accuracy-in-algorithmic-trading#:~:text=Traditional%20algorithms%20can
%20calculate%20risk,data%2C%20helping%20traders%20avoid%20losses.
 https://www.sbisecurities.in/blog/power-of-ai-in-stock-market-analysis
https://www.shareindia.com/knowledge-center/algo/ai-in-algo-
trading#:~:text=The%20algorithm%20is%20quick%20to,in%20the%20best
%20possible%20manner.
https://www.utradealgos.com/blog/what-role-does-ai-play-in-algorithmic-
trading#:~:text=Benefits%20of%20AI%20in%20Algorithmic%20Trading,-
Increased%20Efficiency%3A%20AI&text=Enhanced%20Precision%3A
%20AI%20algorithms%20analyse,and%20data%2Ddriven%20trading
%20decisions.
https://www.imf.org/en/Blogs/Articles/2024/10/15/artificial-intelligence-can-
make-markets-more-efficient-and-more-volatile
https://alpaca.markets/learn/the-role-of-artificial-intelligence-in-algorithmic-
trading
https://therecursive.com/hottest-trends-and-predictions-for-ai-in-finance-a-
glimpse-into-2024/
https://hyscaler.com/insights/financial-trading-success-with-ai-innovations/
https://www.datadynamicsinc.com/blog-ai-in-algorithmic-trading-wall-street-
gets-a-robo-advisor-overlord-heres-why-its-a-good-thing/#:~:text=High
%2Dfrequency%20trading%2C%20market%20making,portfolio
%20optimization%2C%20and%20risk%20management.
https://www.prismetric.com/ai-in-stock-trading/#:~:text=For%20instance
%2C%20Goldman%20Sachs%20employs,minimize%20losses%20during
%20volatile%20periods.
https://www.coforge.com/what-we-know/blog/the-future-of-wealth-
management-ai-powered-personalized-financial-planning#:~:text=Betterment
%2C%20a%20U.S.%20based%20financial,educational%20content%20to
%20guide%20investors.
Risk evaluations:
https://insightss.co/blogs/role-of-artificial-intelligence-in-financial-risk/
#:~:text=AI%20Models%20for%20Risk%20Prediction,-Artificial
%20Intelligence%20models&text=Machine%20learning%20models
%20specifically%20designed,creditworthiness%20of%20individuals%20and
%20entities.
https://appinventiv.com/blog/ai-in-risk-management/#:~:text=By
%20employing%20the%20use%20of,trends%20to%20forecast%20market
%20shifts.
https://www.imf.org/en/Topics/Financial-Integrity/amlcft
https://www.oracle.com/in/financial-services/aml-ai/#:~:text=By%20replacing
%20rules%2Dbased%20software,their%20number%20of%20false
%20positives.
https://www.avenga.com/magazine/integrating-ai-for-smarter-risk-assessment/
https://ideausher.com/blog/ai-in-portfolio-management/
Personal finance tools:
https://sageuniversity.edu.in/blogs/transforming-finances-money-
management-with-artificial-intelligence#:~:text=AI%20Assistance%20in
%20Personal%20Finance%20Management&text=Artificial%20intelligence
%20offers%20personalized%20advice,their%20spending%20and%20savings
%20patterns.
https://www.gartner.com/en/newsroom/press-releases/2024-09-11-gartner-
survey-shows-58-percent-of-finance-functions-use-ai-in-2024#:~:text=The
%20adoption%20of%20finance%20AI,a%20survey%20by%20Gartner%2C
%20Inc.
https://www.researchgate.net/publication/382679575_AI-
Driven_Personal_Finance_Management_Revolutionizing_Budgeting_and_Fin
ancial_Planning
https://trendsresearch.org/insight/smart-finance-how-ai-is-shaping-the-future-
of-budgeting-and-investing/#:~:text=Personalized%20Budgeting%3A%20AI
%2DDriven%20Financial%20Planning&text=%5B8%5D%20This
%20customization%20enhances%20the,with%20their%20specific
%20financial%20objectives.
https://www.thebusinessresearchcompany.com/report/ai-powered-personal-
finance-management-global-market-report
 https://www.datrics.ai/articles/the-essentials-of-ai-based-credit-
scoring#:~:text=Once%20the%20machine%20learning%20model,that
%20predicts%20the%20applicant's%20creditworthiness.
https://www.lseg.com/en/insights/data-analytics/how-might-ai-impact-
investment-management#:~:text=Product%20complexity%20may%20also
%20rise,only%20an%20LLM%20can%20decipher.
https://redresscompliance.com/your-money-how-ai-transforms-personal-
finance-management/#:~:text=for%20individuals%20worldwide.-,Debt
%20Management%20and%20Credit%20Optimization,to%20improve
%20their%20financial%20health.
Case studies:
PayPal on fraud detection:
https://www.paypal.com/us/brc/article/payment-fraud-detection-
machine-learning
https://www.linkedin.com/pulse/ai-case-study-saturday-fraud-
detection-finance-alastair-aaefe#:~:text=The%20AI
%20Solution&text=Key%20components%20of%20PayPal's
%20AI,user%20behaviour%20to%20detect%20anomalies.
https://indiaai.gov.in/article/ai-and-data-analytics-the-driving-
force-behind-paypal-s-risk-management-transformation
https://aimresearch.co/market-industry/why-paypal-is-
aggressively-hiring-in-ai
https://www.americanbanker.com/payments/news/paypal-leans-
on-ai-to-improve-its-customer-relationships#:~:text=PayPal
%20faces%20competition%20from%20other,following%20a
%20two%2Dyear%20correction%20.
https://medium.com/paypal-tech/transforming-paypals-ai-
culture-to-enable-intelligent-business-decisions-b90438b96cd6
 https://emerj.com/ai-sector-overviews/artificial-intelligence-at-
paypal/
https://economictimes.indiatimes.com/tech/technology/paypal-to-
launch-ai-based-products-as-new-ceo-aims-to-revive-share-
price/articleshow/107169681.cms
Wealthfront and Betterment on personal finance
https://www.coforge.com/what-we-know/blog/the-future-of-wealth-
management-ai-powered-personalized-financial-planning#:~:text=Betterment
%2C%20a%20U.S.%20based%20financial,educational%20content%20to
%20guide%20investors.
https://www.linkedin.com/pulse/wealthfronts-automated-investing-
methodology-using-ai-hossein-kazemi-fcvye
https://buddyxtheme.com/best-ai-tools-for-personal-finance/
#:~:text=Betterment%20is%20a%20well%2Dknown,rebalancing%20and
%20tax%2Dloss%20harvesting.
https://rtslabs.com/future-of-ai-and-wealth-management#:~:text=Wealthfront
%2C%20a%20fintech%20company%20specializing,to%20create
%20personalized%20financial%20plans.
Blackrock’s AI algorithms for trading:
https://www.blackrock.com/us/individual/insights/ai-investing
Challenges and considerations:
Bias in AI:
https://www.linkedin.com/pulse/overcoming-bias-ai-finance-industry-monica-
mccoy-3iite#:~:text=Bias%20in%20AI%20refers%20to,overall%20integrity
%20of%20financial%20systems.
https://viterbischool.usc.edu/news/2022/05/thats-just-common-sense-usc-
researchers-find-bias-in-up-to-38-6-of-facts-used-by-ai/#:~:text=Just
%20Common%20Sense'.-,USC%20researchers%20find%20bias%20in
%20up%20to,of%20'facts'%20used%20by%20AI&text=A%20team%20of
%20researchers%20from,found%20that%20it%20wasn't.
https://www.diagnosticimaging.com/view/meta-analysis-high-risk-of-bias-83-
percent-of-ai-neuroimaging-models-psychiatric-diagnosis
Data security and privacy concerns:
https://www.scalefocus.com/blog/ai-in-the-banking-sector-risks-and-
challenges#:~:text=AI%20in%20banking%2C%20therefore%2C
%20raises,to%20the%20AI%20integration%20process.
https://www.bccpa.ca/news-events/cpabc-newsroom/2024/june/ai-risks-in-
accounting-and-finance/#:~:text=Some%20of%20the%20typical
%20privacy,use%20data%20for%20other%20purposes.
https://www.fsb.org/2024/11/the-financial-stability-implications-of-artificial-
intelligence/#:~:text=GenAI%20also%20increases%20the
%20potential,behaviour%20that%20harms%20financial%20stability.
https://www.ecb.europa.eu/press/financial-stability-publications/fsr/special/
html/ecb.fsrart202405_02~58c3ce5246.en.html#:~:text=However%2C
%20hallucination%2C%20algorithmic%20bias%20and,or%20even
%20disorderly%20market%20moves.
Job displacement:
https://www.imf.org/en/Blogs/Articles/2024/01/14/ai-will-transform-the-
global-economy-lets-make-sure-it-benefits-humanity
https://www.imf.org/en/Blogs/Articles/2024/01/14/ai-will-transform-the-
global-economy-lets-make-sure-it-benefits-humanity
https://www.imf.org/en/Blogs/Articles/2024/01/14/ai-will-transform-the-
global-economy-lets-make-sure-it-benefits-humanity#:~:text=If%20AI
%20significantly%20complements%20higher,these%20phenomena%20could
%20exacerbate%20inequality.