Where to Begin When Looking for an Accounting AI Solution
After possibly months or years of stalled digital transformation conversations and further delays due to COVID-19, your manager or one of your firm’s executives has finally given you the green light to start pursuing artificial intelligence (AI) and machine learning (ML) accounting solutions.
It’s great news that will allow your business to remain competitive, but where do you start? Here are a few tips to consider when looking to evaluate what AI can bring to your outsourced accounting practice.
Is it Really AI?
Let’s clarify what AI is. It’s machine technology that’s able to solve for certain tasks typically performed by humans, which tend to be related to classification and prediction. On an average day, you probably interact with AI technologies like voice-activated assistants Alexa and Siri; email spam filters that can separate the junk from regular email; and Lyft and Uber, which use AI to figure out when your ride will arrive.
But most importantly, AI learns from experience and can get smarter on its own.
You will no doubt find tons of solution providers claiming that their technology can automate various aspects of accounting processes. In reality, many of these technologies aren’t AI but rather Optical Character Recognition (OCR), which is software that recognizes text from a document, like an invoice. Both OCR and AI are used to automate tasks, but OCR can only function based on rules that humans have programmed it with. With AI, however, the technology can think for itself through continuous exposure to data.
If you take time to look under the hood, the most valuable question you need to ask is, “Does your AI learn on its own?” If the answer is “No,” then it’s not AI, and the system will never be able to do more than what you have to dedicate time and effort programming it to do.
The Proof is in the Data
AI is only as smart as the data it’s been exposed to. The quality of the data is responsible for making the technology as accurate as possible, so this is no small concern. It’s the same as eating a meal - if you “fuel” AI with a lot of clean data, like eating your fruits and vegetables, it will get smarter. If you feed it data that’s not clean, it’s the same as eating fast food - you get out of the AI what you put in. For AI that tackles invoice processing, it’s important to understand things like:
- How many financial documents and transactions has the AI trained on? (If it’s in the hundreds of millions, or has been trained for years, that’s a good sign)
- Have actual auditors audited the data?
- Does the data meet strict privacy and regulation laws?
- Is the data set clean?
All these elements contribute to the intelligence of the AI.
Plays Nice with Others
True AI is similar to other technologies, like the cloud, in that it is agnostic and can be used with multiple other platforms. As you start evaluating AI accounting vendors, you’ll want to look for a solution that allows you to switch easily between systems and not have to overly rely on a single platform.
Like a member of your , AI should be able to work with any technology. Your should have some level of ownership of the AI since it would be developed from your client invoice data and the associated interactions. The AI shouldn’t be “trapped” or “locked” into a system you don’t have full control over. That would be like hiring a new staff accountant and only giving them access to one of the eight platforms they need to do their job.
Don’t Outsource Outsourced Accounting
In the rush and excitement to launch your accounting firm into the future, you might find that you’ve overlooked a critical question: what exactly are you going to use AI for? Sure, you’re probably going to use it to automate your processes, but which processes? And what will the anticipated ROI be?
AI, like any other technology, is not a miracle cure or a magic wand. Pick a specific use case your firm is trying to solve for - that way, you have a clear focus when selecting an accounting AI provider.
For example, if an AI platform claims to be able to run your entire business for you, that should raise some red flags right away.
When you start outsourcing your and , the first thing that you lose is control. Control over things like:
- The security of your clients’ data. Do you know how many third parties have handled it? Does the data live in the U.S. or somewhere else that has more lax regulations?
- The accuracy of the accounting AI. What is the error rate for the AI, and how can you tell if you can’t see for yourself?
- The intelligence of the AI. Is the AI operating on its own, or is it overseen by another bookkeeper who may not even have your team’s skills?
- Your revenue. If your clients interact directly with your accounting AI vendor and have to turn over all your books, that’s going to start eating into your margins. Where does that leave your business?
If you’re looking to add AI to your organization, you’re not looking to do so because you want to outsource your operations entirely and put yourself out of business. You’re interested in accounting AI o make your business more efficient and accurate so that you build client trust and have time to scale your organization.
AI automation should be something that your team can use and control internally to do tasks like processing invoices better.. It should not be something that competes with your business model. Do you really want to outsource your outsourced accounting department? Or do you want to fix it so you can keep all revenue in house?
The choice comes down to hiring more headcount or finding the right technology partner to enhance your team’s capabilities. Like hiring a new assistant, your accounting AI should continuously get smarter so it can function as the new, little genius team member who always remembers what it’s been taught, never gets tired, whose talent you never lose to other opportunities, and who you manage directly.
The options for accounting AI are constantly growing and can get pretty confusing very quickly.
The challenge is not just figuring out what technology to source, but also figuring out what problem your organization needs to address first. Without that key step, whatever AI platform you choose, no matter how advanced, may prove to be utterly useless for your purposes.