Generative AI has exploded, and AI hype is everywhere. The internet is awash in a whole new acronym soup, and the urgency to get aboard is palpable. But what are the AI use cases in Finance and Accounting? How will the office of the CFO be required to weigh in on budget decisions for the adoption of these technologies org wide? In this article we will unpack the lingo, and look at the impact of AI for the CFO.
Breaking down AI terminology
AI, Artificial Intelligence, is a sweeping term that extends forward into amazing predictive and interpretive abilities of machine learning. It is also being extended backward to include older digital decisioning processes.
Lets start by unpacking the acronyms and defining the terminology and its applications for finance leaders.
ML, or Machine Learning enables software robots to get faster and more accurate over time. It “practices” making determinations and receives feedback on its accuracy from humans. By logging its successes and corrections the model “learns” to be more accurate over time. ML Models are at the heart of AI digital decision making.
An ML model is trained on sample data, utility bills from hundreds of companies is a document processing example. Then when a new utility bill is processed by the software, it creates a confidence score. The software “decides” how sure it is that this document follows the patterns of other documents it has processed.
In the training phase a human team member will audit the software’s outputs. They will validate correct outcomes and correct errors. The more feedback the model gets the more confident it becomes in its ability to read your utility bills correctly every time.
GPT – Generative Pre-trained Transformers are a form of Generative AI. They have been trained on millions of examples of the way people communicate. It can be used to create images, sounds or text. They use Large Language Models (LLM’s) to predict which words might go together to generate an appropriate response to a question.
ChatGPT is a conversational AI by Open AI. It has been programed to carry on a conversation, asking and answering questions, with some ability to “remember” the past exchanges in the same thread.
For creative challenges and general research they can be fantastic at providing large amounts of information. However, for factually based queries one must exercise caution. The software is designed to generate “something that answers the question” there is no internal fact checking. The goal was to create “something that sounds right”. Which can be dangerous when factual accuracy is required.
Private Secure ChatGPT solutions can be created using selected LLM’s and training the AI on company specific data to increase the fact base that the AI has to pull from. This increases the accuracy of generated outputs, but human review is still needed to confirm that the content is factual.
An LLM is a Large Language Model. It stores millions of data points, facts, phrases, and common uses of language in a vector database. These machine learning models are the heart of any Generative AI software. The AI can then perform Natural Language Processing (NLP) tasks, like describing steps in a task, or writing an email.
Each model must be trained on sample data. What data you train the model with dramatically effects your outcomes. In Corporate settings if you train an LLM with company specific data. Text from your company website can be used to provide relevant information to customers about products or services through a chat bot.
Train the LLM on your outbound emails, the AI could be asked to write emails in the company voice. Include company protocols and FAQ’s and your staff can query the chat bot for questions like, “How many days do we get off for Christmas Holiday?” rather than emailing the HR department.
OCR stands for Optical Character recognition. It is the ability of software robots to read printed or handwritten characters on documents and create a transcript of their contents. Intelligent Document Processing (IDP) applies AI machine learning models that allow your digital workers to Understand and Classify data extracted from documents.
When these AI tools are combined with RPA – Robotic Process Automation, they create end-to-end Intelligent Automation Solutions for Document processing.
Jumpstart your AI implementation with a free use case assessment from one of our AI Experts
Is RPA AI?
Strictly speaking no. RPA – Robotic Process Automation is achieved by creating a set of instructions for the software robot to follow. They must adhere to business rules set for them and don’t do any of their own thinking. In the modern conversation they could be looked at “dumb robotz.” But what that actually means is these reliable worker bees will do exactly what is asked of them every time exactly the way they were told to do it.
AI involves some kind of Machine Learning, and decision making. Digital Decisioning AI has been around for a long time. This could look like generating a confidence score for how well a document was “understood” by the computer and how well it was able to catalogue the entries. It is a wizard at finding patterns in business data, and it’s what powers how Netflix knows what video to show you next.
The Generative AI explosion of the past year has made Conversational AI and Natural Language Processing available to the public. WonderBotz has been working with these technologies for years, the internet has now caught up and is creating a wave of FOMO that cannot be ignored.
Gen AI is a rising tide that will raise all botz– Cathy Tornbohm – Gartner Analyst
RPA + AI = IA
Traditional RPA is getting smarter, that’s why we call it Intelligent Automation.
Combining the two technologies creates a best of all worlds situation. The safe reliable automations that make business run, are enhanced with the ability to make decisions and take on more complex tasks.
Intelligent Automation Solutions opens the doors to new use cases that were too hard to program without AI to find the critical path through the storm of data.
Generative AI promises another wealth of employee productivity jumps and customer experience improvements. As with any new technology it comes with its own risks and Cybersecurity is top of mind for ChatGPT usage. However, with proper data governance, enterprises can offer their staff and customers these tools and maintain data security. With the right guideposts in place, your new Smart Robotz will be just as safe, just as reliable, and just as compliant as the old ones.
AI for the CFO
We have unpacked the alphabet soup and sorted through the hype, and AI is here to stay. With mandates from many companies to implement emerging technologies, budget discussions will fall to the office of CFO. Jack McCullough from the CFO Leadership Council cites the ability to participate in technology strategy as a “must have for CFOs”.
In some organizations the CISO’s report to the CFO, brining cybersecurity and cyber insurance into the technology conversation. New ESG reporting requirements for scope 2 and 3 emissions, extracted from invoicing and financial reports, are reporting into the CFO as well.
As the role of the CFO evolves beyond finance and accounting into technology strategy for the organization, AI for the CFO becomes a must have. Let WonderBotz help you and your organization navigate these new demands.
Ready to add AI-Powered Intelligent Automation to Your Organization?
As an Automation firm that specializes in Finance and Accounting solutions, WonderBotz is excited about the way that AI added to RPA is opening new use cases. We have been implementing pioneering technology in the automation field since our inception. As early adopters we have had the time to sort out the hype of the current GenAI craze. We are focusing on helping enterprises operationalize AI technology and capitalize on the huge advancements in end-to-end automation capability.