Mosaic | External Link
Financial models often resemble a house of cards — a delicate balancing act of assumptions working to prop up the business. You wouldn’t expect modeling to be a prime use case for AI. But we’re closer to the reality of AI financial modeling and forecasting than many would think. Here’s what to expect.
The promise of generative AI is that it can take tasks and concepts which are complex, time consuming and data heavy, and make them fast and simple. That’s good news for finance teams looking to improve their financial modeling and forecasting capabilities, because those are most definitely tasks that match that description.
Traditional financial modeling and business forecasting has always involved a huge amount of manual work. Financial data needs to be collated and checked for errors, models then need to be manually built from scratch or templates, which introduces a whole host of new potential problems.
Once they’re done, the process of updating the models requires more time again, and of course there’s that wonderful statistic from the European Spreadsheets Risk Group (oh yes, it’s real) which estimates that over 90% of spreadsheets contain an error.
When you’re in charge of financial forecasting for a large company, a simple error could cost millions.
AI can help make financial modeling and forecasting much quicker and easier, allowing you to make adjustments on the fly, use natural language to build complex reports and even have AI identify trends, risks and opportunities that you might miss. All while reducing the risk to your business, through its ability to spot anomalies and potential errors.
Sounds pretty good doesn’t it?
The Benefits of Using AI for Financial Models and Forecasts
The term AI has become a bit buzzy lately. It’s no surprise because the potential for the tech really is huge. But because the word has been filled with so much hype, it can be easy to lose its true meaning. In the context of FP&A modeling and forecasting, it’s probably useful to think of AI as ‘automation.’
That’s effectively how it provides benefits to this area of finance, by taking complex tasks that are usually done by humans, and automating them. Now obviously those automations can have multiple steps and be highly complex, but in essence that’s the key benefit of using AI for financial modeling and forecasting.
More specifically, some of the key benefits of using AI or automation technology are:
| Streamlines Workflows |
| By replacing time-consuming data entry and model building tasks, you can save hours every week in time by automating parts of the existing process. And with an AI-driven chat function, like Mosaic’s Arc AI, this can be integrated into almost any part of your process. |
| Reduces Errors |
| It’s incredibly easy to ‘fat finger’ the wrong number in a spreadsheet, or copy and paste the wrong formula. AI doesn’t have that problem, meaning errors can be reduced significantly or even eliminated altogether. |
| Lowers Technical Proficiency Requirement |
| Because AI can speed up the forecasting process so much, it means your team has the ability to drastically improve their scenario planning capabilities. Rather than needing hours to model a single scenario, you can now iterate many times over based on different assumptions, and even create scenarios on the fly as the landscape changes. |
| Increases Collaboration Possibilities |
| Building a financial model from scratch requires a skilled financial analyst. However, through the use of AI large language models (LLMs), the formulas and spreadsheets can be replaced with a normal language prompt. That means anyone in your team can run financial forecasts and build models, not just the technical analysts. Of course, it’s important that the outputs are verified by the right people to make sure they’re accurate. |
| Improves Data Volume Management |
| There are practical limits on the size of the datasets that can be handled by spreadsheets. For example, trying to integrate Stripe data for a multinational company could mean millions of lines of data, which is simply too much for Excel or Google Sheets. By using AI, this data can be collated and organized without the need to import it to a spreadsheet. |
How AI is Used in Financial Modeling and Forecasting
In a recent round table discussion on AI in finance, Rob Matthews from Spiff put it best when he said that AI should be used to do “the simplest tasks that suck the most.”
And while that’s funny, it’s actually a very clear message that you and your team can use to decide which parts of your workflow can likely be replaced by AI and automation. Because one of the best ways in which AI can be used in financial modeling and forecasting right now is in organizing your data in the first place.
Collating financial information from multiple sources manually can, in many cases, be replaced by AI, which can combine and categorize large datasets to give you a starting point for your analysis.
Not only that, but with machine learning models and predictive analytics, it can also help identify anomalies in that data, as well as trends, opportunities and risks. Again, these aren’t necessarily all going to be relevant for you, but it provides a prompt to your human analysts to apply their strategic thinking to the numbers.
As Mosaic, we see AI as becoming a combination of a data administrator, financial analyst and data modeler, giving you clean data and baseline models, from which your human team can fine tune and iterate.
Put simply, it will allow your FP&A team to spend less time on the ‘What’ of financial modeling and forecasting, and more time on the ‘Why.’
Security Considerations when Using AI for Financial Forecasting
But with that said, there are definitely some things to keep in mind when it comes to security. AI models like ChatGPT work by being ‘trained’ on massive data sets. If you ask a question like, “How do I hang a picture on the wall?” it will provide you with what it determines to be the most relevant response based on the data points it’s been trained on.
That could be anything from low quality affiliate blog posts to high domain authority websites to Reddit posts or even Tweets.
The important thing to note here is that any data that is input into the model by its users can be added to this training dataset. It’s one of the reasons why ChatGPT, for example, gives you the ability to give a thumbs up or a thumbs down to the answers it gives.
Long story short, that means that any information you input into an AI model can potentially be scraped by that model. It goes without saying that you should therefore be very careful about any data that you wouldn’t want third parties to get their hands on.
There are, however, tech solutions being worked on that utilize the full capability of AI without compromising data security.
One example is to have your data remain within your trusted ecosystem, and then using an LLM as a sort of ‘translator’ to turn your natural language question into a technical request which is then relayed to your dataset. The LLM never actually sees your data, just the question you’re asking of it.
The important thing is to ask the question, and understand how your data is treated by any AI technology you want to implement into your business.
How to Implement AI in Financial Modeling
So there are some very clear benefits to implementing AI into your company’s financial modeling. But how exactly do you go about implementing it into your business? There are some key steps to consider.
Identify Potential Use Cases
The first step is to gain an understanding of how you might utilize AI within your finance function. As Rob from Spiff says, thinking about the simple tasks that suck the most can be a good place to start. You’re not going to be able to see every possible AI integration right now, but highlighting some key ones can help you with the next step.
Find the Right Solution
With an idea of the kind of functionality you might want, you can look for financial modeling software that does those things. Maybe you want an AI assistant that can pull insights from your existing data, or build models based on different scenarios, such as a reduction in marketing spend (Mosaic’s Arc AI integration does all that and much more, by the way).
Use the Tool
Once you’ve ‘onboarded’ your new AI assistant, it’s important that you and your team constantly think of ways to utilize it. With any task you plan to do yourself, it’s worth trying with the AI assistant to see how it approaches the task and how you find the quality of the end result.
It’s only by trying a wide range of different tasks that you’ll begin to understand where it offers the biggest benefits. Not only that, but if using it becomes part of your workflow, you’re more likely to uncover use cases you had originally thought out.
Trust, but Verify
Mosaic’s approach to AI is very much ‘Trust, but verify.’ Our co-founder Brian Campbell makes the excellent point that even for new human hires to your finance team, you verify their work as they get embedded in your team.
Over time and once the technology (or real person) becomes familiar with your processes, you can trust more and verify less.
Mosaic: Pioneering AI-based Financial Modeling
Mosaic has always been at the forefront of automation and moving the FP&A function into a more strategic part of the business. Focusing less on tallying up the numbers from the past and spending more time on helping guide the financial decisions of the future.
The integration of our new Arc AI into the Mosaic platform serves to exponentially increase the capabilities of finance teams, allowing them to shift more and more time away from manual, low value tasks and more into the high value analysis and strategic functions.
Mosaic has always provided a single source of truth for your company’s data, and the ability to create complex financial modeling and forecasting outputs. The difference now is the way your team will be able to interact with your data.
It can be as simple as being able to ask ‘Why?’ or ‘What if?’
Arc AI is launching in early 2024, and will give users a chat interface, allowing them to interrogate the numbers like never before. For example, say your ARR has increased substantially over the last quarter, and you’re not sure where this growth has come from.
Rather than spending hours digging into the data yourself, simply ask Arc “Why has our ARR increased so much this quarter?” and it will analyze the numbers and tell you.
Or perhaps you want to do some SaaS revenue forecasting with the introduction of a new premium price point. Well, you can just Arc to put it together for you.
When it comes to financial modeling, you can use Arc to model changes to advertising spend, headcount, pricing and just about anything else you can imagine. And because it can be done at the literal click of a button, you can run models immediately when the business landscape changes.
In short, it means your finance team can put all of their time, expertise and energy into strategic decision making, which is the ultimate value add to the broader company.
To see how Mosaic and our Arc AI can help elevate your finance function, book a demo today.