If the idea that AI could change the way we work was merely a whisper two years ago, then today, it’s a full-on siren.
In 2024, artificial intelligence & automation are cementing themselves as major investment areas for businesses: SAP Concur’s 2024 CFO Insights Survey shows that 51% of CFOs are investing in AI this year as opposed to 15% in August 2023.
That said, the AI landscape is increasingly complex: in the same SAP Concur survey, 58% of finance leaders admitted to understanding very little about AI in finance. With the amount of terms in the space, you might confuse the AI conversations in your midst for just some things the kids are saying these days.
To make it easier for you, we’ve parsed out the relevant information to keep you in the know.
AI Acronyms – Alphabet Soup of AI
Breaking into AI begins with breaking down some of the lingo within the field – namely, the slew of acronyms that you will or maybe already have interacted with.
NLP
NLPs, or natural language processors, refer to the vast array of tools that are focused on interpreting and understanding human language. NLPs are trained to understand grammar, syntax, semantics, and even nuances like sarcasm.
Siri, Apple’s famous phone assistant, is a classic example of an NLP where spoken commands are interpreted and translated into actions.
OCR
Optical character recognition (OCR) is an AI process that converts the text within PDFs, written notes, images, and more into machine-readable text. If you’ve ever watched software convert your receipts into readable and editable text in seconds, you can thank OCR technology for that.
As such, it’s nearly unthinkable that any modern accounting team would not be using some form of OCR – if not, doing so could be the first step toward cutting down on your close. For month-end bank reconciliations in Numeric, OCR scans the entirety of bank statements, identifies ending balances, and pulls in that balance to compare against the trial balance from the GL.
LLM
If you’ve ever written a prompt in ChatGPT, then you are already familiar with LLMs. As perhaps the most talked-about area of artificial intelligence today, LLMs – large language models – can be viewed essentially as an evolution of NLPs.
Trained on billions of pages of data, these models can not only interpret human text, but also generate human-like content. In turn, their use cases are seemingly endless: writing a script for a potential new episode of Seinfeld, re-teaching you forgotten high school calculus concepts, planning an itinerary for a week-long getaway to Cabo, and plenty more.
Common LLMs
![](https://cdn.prod.website-files.com/62fd7949a6c5ed5af6dd8f09/66a9826b13f975e2a5fa2f88_66987abb8ce329401ddc840c_1.png)
- ChatGPT: Best all-around and most accessible LLM
- A household name when discussing AI, ChatGPT was built by OpenAI and initially released in late 2022. Great for quick text summaries, a Google search equivalent, or spinning up a quick image.
![](https://cdn.prod.website-files.com/62fd7949a6c5ed5af6dd8f09/66a9826b13f975e2a5fa2f82_669880bf5a8d22986e0df00f_chatgpt.png)
- Claude: Best for writing-focused tasks
- Considered ChatGPT’s rival and equipped with a more personable tone of voice in its responses, Claude was released in early 2023, courtesy of Anthropic.
![](https://cdn.prod.website-files.com/62fd7949a6c5ed5af6dd8f09/66a9826b13f975e2a5fa2f7f_66987acd8ce329401ddc8e83_4.png)
- Gemini: Best for multi-media content generation, i.e. text-to-image
- In contrast to other LLMs that are only trained on text, Google touts Gemini (initally Bard, released 2023) as being multimodal, having been trained on text, videos, computer code, and images.
![](https://cdn.prod.website-files.com/62fd7949a6c5ed5af6dd8f09/66a9826b13f975e2a5fa2f90_66987ad801dd727b046b9de1_3.png)
- Llama: Best for coding tasks
- Meta’s first released Llama in February 2023. While Llama 3 has a more explicit coding focus, one can use chatbot Meta AI for text-based conversations.
What LLM should I use?
For the vast majority of use cases, any of the models will work just fine. For accounting-specific questions, customers can use Numeric’s Technical Accounting AI to provide answers across accounting standards (IFRS, GAAP, etc.), SEC filings, and Big 4 literature.
AI in the Wild – what are Embedded vs. All-Purpose tools?
While LLMs continue to grab the most attention in the AI space, you’ve probably also seen software products you already use touting their new AI capabilities. If you’re at all wondering how those differ, let’s make a distinction.
We can think of tools like ChatGPT and Claude, or even image-making service DALL-E as more general purpose AI products. In short, they are perfect for quick on-the-fly requests and can quickly generate text or visuals. In doing so, this genre of AI often replaces the traditional Google search with a more robust and customizable search function.
Contrast this with embedded workflow AI. Embedded AI sits within an existing platform and assists you as you move through your workflow. Numeric employs embedded AI across its platform, particularly with its flux analysis AI writer – here, teams can have the AI writer generate variance explanations that identify key drivers of change pulled from the data that’s in their GL. The flux writer isn’t built to perform nearly any of the tasks ChatGPT can: it’s meant specifically to serve Numeric customers with one element of their month-end workflow.
Succinctly stated, embedded AI is powered in the same way as all-purpose AI products with LLMs, but with a significantly smaller scope and a defined purpose for its parent platform.
What tasks are best suited for AI?
If your only experiences with AI have come in the form of crudely rendered hands with six fingers or less-than-stellar movie scripts, then you might be dubious of AI’s performing power. In fact, one might argue that these rough outputs are what have many people skeptical of AI’s long-term potential.
Yet when applied to the right tasks, AI can actually produce some impressive work – remember that these are high-octane computing systems. To get the most out of them, you just have to know where in your workflow AI could work best.
For many, these tasks will fall into the following categories:
Scanning
The ability to convert huge, printed documents into readable text is nearly a modern marvel. Because of OCR, people should rarely, if ever, spend time rewriting or recreating physical documents – as such, leave the printed leases, bank statements, and other physicals in your repertoire in the hands of document-reading programs, not someone on your team.
Generating
If you wanted a 2000-word draft about the history and nuance of ASC 606, AI could do it in roughly the same amount of time it takes for you to walk from your desk to the door. Seriously.
Are these responses perfect? Not necessarily – they may require editing for tone and voice – but by and large, they are perfect for getting the first drafts of your writing accomplished.
Polishing
Almost inverse to generating, AI is excellent at fixing up your own writing or code as well. If you’re unsure of grammar and format or you just want your writing to be more concise, a quick AI writing check can ensure that your writing is as tight and well-worded as possible.
Summarizing
AI systems can crunch numbers and words at far greater speeds than humans. Say you asked an AI bot to summarize the changes driving month over month variance in a balance sheet account: what would take a human potentially hours to identify takes AI seconds.
See how accounting teams use AI in the month-end close
Is Automation the Same Thing as AI?
One final acronym sits beside the rest – that is RPA, or robotic process automation. Why? Well, technically speaking, RPA is not truly a form of artificial intelligence.
The two are similar because they each rely on pattern recognition to imitate some human element – automation mimics the way humans might do a task while AI mimics human thought. However, automation only works to accomplish highly structured and logical tasks – if AI can generate creative outputs that we maybe can’t create ourselves (writing a Seinfeld episode), then automation tackles tasks we very well could do but with much more velocity and at scale (copy-and-pasting values).
So, is RPA inferior to AI? Not at all! Instead, automation is perfect for a number of use cases, such as addressing the repetitive drudge work that takes us away from more important tasks.
Let’s think about this in the context of the month-end close. Maybe every month, you find yourself going across different portals, downloading all your necessary workpapers, and then saving them to one folder. Difficult? Not really. Tedious? Absolutely.
By setting up an automated process or using various RPA software, you can take this menial labor out of your hands. The same goes for doing file conversions – instead of having to go in and manually convert files to the format you need them to be in, automation can do this for you much faster and many times over without getting exhausted.
Think of automation and AI as productivity partners – one gives you time back by performing your tasks more efficiently and at scale; the other saves you time by filling in the blanks when your mind either can’t or doesn’t need to.
How do AI + automation help accountants?
As Numeric CEO Parker Gilbert explains, he feels pretty strongly that AI won’t replace the roles of accountants. In fact, he actually believes that leveraging AI and automation technologies will serve as the best solutions for addressing the prominent issues accountants face right now, including the ongoing accountant shortage.
How so? For the overworked and understaffed teams typical in many accounting departments, AI offers a solution to eliminate the manual tasks that frequently slow down progress.
Picture having a CPA assistant at your disposal to tackle mundane tasks like drafting the right credits and debits to handle a fixed asset disposal. Imagine a reality where instead of poring over endless spreadsheets to catch accrual errors or misallocated expenses, AI can swiftly analyze variations and in the same breath, write first-draft flux explanations for you.
Consider the example of policy generation: traditionally a labor-intensive process involving extensive research, nuanced adjustments, and writing detailed policies. With Numeric's AI policy memo generator, you provide specific business context, and AI transforms it into a polished accounting policy ready for your review.
No more hours wasted searching for GAAP answers online; simply ask AI and receive immediate responses.
As AI and automation rewrite the accounting playbook, embracing these technologies is becoming increasingly essential for staying competitive. By leveraging AI-powered tools and automated processes, accountants can free themselves from mundane tasks, focus on higher-value work, and ultimately drive better financial decisions for their organizations.