AI from a business perspective

With its broad range of capabilities and levels of independent thinking, AI applications are able to play a role in just about all aspects of society, especially in business. The aim of this article is to expand on how businesses utilise AI and its effects on factors like productivity, costs, etc.

The mass availability, as well as the quantity of data that businesses have, offer opportunities to provide key insights to leadership teams about potential business decisions. Artificial Intelligence is a step-up from the traditional data analysis methods that currently exist. AI applications may use the very same models used by traditional data analysis, the distinguishing difference being that AI is capable of presenting the analysis of data with more customization and in a broader context. AI solutions can output data analysis (including recommendations to the research analyst) that may not have been apparent or even considered. With the ever evolving processing power of technology, in which researchers have access to hardware, allows for the processing of immense amounts of data not previously possible.

‍

Differences Between Machine Learning and AI for Data Analysis

Machine Learning models (a subset of Artificial Narrow Intelligence) were designed for data analysis and have existed for over 50 years. At this point however, given all the technological advancement, these can now be more efficiently utilised. Machine Learning solutions will conduct data analysis and process to the requirements of the researchers. They do not however provide the additional insight that an Artificial General Intelligence (AGI) could provide. ML solutions can only offer recommendations limited to what it can process from actual data, whereas an AGI can access a knowledge base from sources beyond the data. This access to external knowledge bases allow AGIs to provide unique approaches to problems that researchers may encounter (ones they did not possibly even anticipate). Businesses that have quite a rigid framework dictating how they go about their data processing and analysis should therefore opt for using machine learning models over AGI solutions.

Microsoft’s Azure Machine Learning allows for the building, testing and calibrating of custom AI models using custom data. The skill and experience required to assemble a team capable of developing a machine learning model for a business (from scratch) could be quite a strain on a company, especially if there is no AI experience within their software development team.

Automated Machine Learning (AutoML) automates and eliminates the manual processes in machine learning model development. It allows for less experienced developers to have access to machine learning by taking away the complexities of developing a model and putting more focus on the training data set as well as the testing and evaluation of a model.

In general when choosing an AutoML model, the service provider will run multiple models on the data and grade the output based on the desired parameters from the user. Ultimately the model that meets the user’s requirements is then selected. Google’s Vertex AI offers an AutoML option with in-depth instruction on how to use the feature for generating, testing and deploying models. Microsoft’s Azure Machine Learning also offers an AutoML option which offers users the choice between a code or no-code experience.

Using a Machine Learning model on its own can be useful for accomplishing desired tasks. In business there is however a requirement for more functionality like offering detailed insights on model output, or offering assistance to users via a chatbot or co-pilot.

Most solutions offered by AI providers comprise layers of functionality involving the utilisation of multiple models. In order to make use of multiple models in an efficient way, a stronger AI (AGI) is required. An example of an AI solution that is capable of performing functions across multiple aspects of business is Salesforce’s Einstein AI. Einstein has several variants including fields such as: Sales, Commerce, Customer Service and Marketing. Another example of automation covering multiple facets of business is Creatio’s Studio platform which offers solutions focused on Sales, Customer Service and Financial Services fields.

The above mentioned examples cover specifically tuned solutions, however businesses may opt for a more general solution and then work on leveraging the output to provide precise analytics where desired.

Business Processing using AI

Businesses have quite notable processing requirements for a variety of aspects such as paperwork, business reports, meetings, client call transcripts, etc. Anthropic’s Claude AI provides a solution capable of processing large amounts of information (latest iteration has a context window of 200K tokens) and can provide insights (text and code generation) on just about any topic presented to it communicating in Natural Language.

OpenAI’s ChatGPT is widely available as a Large Language model (LLM) AI solution proving generative text based responses to prompts that are inputted as basic Natural Language (ChatGPT’s latest premium model offers document upload capability) but there are voice and image input capabilities becoming more available with the offered models (DALL-E is OpenAI’s image model that is used in conjunction with ChatGPT and allows for image processing and generation).

Microsoft co-pilot offers a similar text based prompt response system to the above mentioned solutions however it also comes with the added advantage that it is embedded within Microsoft software applications and can execute commands within them. With IoT only increasing in relevance, Information Technology divisions have to ensure businesses are running as efficiently as possible.

AI software engineering solutions such as Devin AI exist to aid software teams in this goal. Devin is capable of autonomously handling full software development projects, from planning to execution, together with an embedded adaptability to the ever changing landscape of code.

Phind AI serves to be an upgrade on traditional search engines by leveraging the contextual and reasoning strength that AI possesses to provide more user focused and relevant retrievals per input prompt.

IBM’s Watson AI comes in several variants, these:

• are used to generate, classify, summarise, extract and perform QA on data
• serves as a data store for generative AI, providing generative AI-powered data insights, real-time analytics and business intelligence, and access to data across hybrid Cloud all in a single integrated console
• governs generative AI models built in watsonx.ai and on third-party platforms such as Amazon Bedrock, Microsoft Azure and OpenAI. Watsonx.governance is used in applications such as credit risk monitoring and evaluation, customer service and supply chain management

How can AI be used in aspects of financial businesses?

Artificial Intelligence used in fintech is quite often implemented in its Machine Learning guise providing solutions in notable functions of financial business. Some of these are:

• Compliance: AI automates the monitoring and reporting requirements thus ensuring regulatory compliance (e.g. Watsonx.governance).
• Credit scoring: The analytical power of AI can be used to analyse customer data such as online activities and behaviours to assess their creditworthiness and make more accurate credit decisions (e.g. Socure is used by institutions like Capital One, Chime and Wells Fargo).
• Customer service: AI-powered personal assistants and chatbots reduce the need for human intervention and provide personalised customer service such as real-time credit approvals, and offer consumers improved fraud protection and cybersecurity.
• Fraud detection: AI bolsters the security of activities such as online banking and credit card transactions. AI is capable of detecting unusual patterns in financial transactions that would not be easily picked up by human analysis and thus aid in preventing financial crime, such as fraud and cyberattacks. (e.g. unsupervised learning to detect anomalies in groupings of data points describing financial behaviour of a subject).
• Loan processing: Through automation of tasks such as risk assessment, credit scoring and document verification, AI streamlines the approval process. It is also geared to better predict and assess loan risks.
• Risk Management: AI can analyse data to help financial organisations assess and manage risks more effectively and create a more secure and stable financial environment. Simudyne offers a platform that allows financial institutions to run stress test analyses and test the waters for market contagion on a large scale. The company offers simulation solutions for risk management as well as environmental, social and governance settings.

Additional AI applications in finance

Some additional applications of AI in finance include aiding consumers in the management of personal finances and portfolios assisting research analysts by providing predictive analytics, risk management (e.g. Derivative Path’s platform helps financial organisations control their derivative portfolios. The company’s cloud-based platform, Derivative Edge, features automated tasks and processes, customizable workflows and sales opportunity management. There are also specific features based on portfolio specifics, for example, organisations using the platform for loan management can expect lender reporting, lender approvals and configurable dashboards) and sentiment analysis.

Who are the role players in implementation?

Building on the understanding of how AI is utilised in the finance industry, some of the various role-players in dealing with AI solutions’ implementations and outputs are:

The auditors and internal control teams are responsible for assessing the performance of the AI systems and conduct audits on these systems to identify issues and monitor compliance.

Chief information officers (CIOs) and chief technology officers (CTOs) make key decisions regarding AI implementation, usage and security.

Developers: AI developers are responsible for the actual designing, implementing and monitoring of the AI solutions.

Ethics and diversity officers guard against bias, ensuring fairness and inclusivity in the use of AI.

Legal teams work with regulators ensuring the compliance of AI applications with relevant laws and industry regulations.

Risk management teams monitor the effectiveness of AI systems used in risk management applications providing a second layer of assurance to the output of the model.

Conclusion

Businesses have many moving parts and each of these parts pose opportunities for the introduction of AI to improve them. Businesses working with large information sources (documents, transcripts etc.) may opt for solutions such as Claude and ChatGPT as they are likely being bottlenecked waiting for employees to read and summarise before taking action. Both solutions can summarise these sources as well as provide high accuracy QA saving  businesses a significant amount of time. In environments where Microsoft software is predominantly used, having access to a chat-bot styled Co-Pilot that assists in information retrieval and insight provision (together with the capability to execute commands based on a prompt) can be advantageous. Devin can be used for the software engineering needs of a business and Phind offers an improvement in search queries for businesses requiring more niche and complex information retrieval.

The handful of AIs mentioned span across multiple branches of a typical business (and don’t even include the field specific applications mentioned here) . Used in a responsible and purposeful way, AI used in a business has the potential to truly re-define what success can look like.