ChatGPT opens floodgates for vendors on AI-powered accounting solutions

Normally it takes at least a few years for a technological advance to work its way through the accounting profession, but ChatGPT stands as a stark exception to this rule, going from near-complete unknown status to practically all anyone can talk about within just a few months.

What has changed is not so much the presence of AI itself — which has been around in various forms for years — but its widespread availability. Before, AI had been largely the domain of giant corporations that had hundreds of millions of dollars to invest in developing their own proprietary systems. Now such power is available to anyone who can pay the OpenAI licensing fee, allowing accounting software vendors to offer novel new products. 

For instance, workflow and client collaboration software company Client Hub recently announced the release of its new Magic Workflow solution, which uses ChatGPT to generate checklists for tasks. Users tell the tool what it is they want to do — like preparing a month-end report for QuickBooks — and Magic Workflow generates a step-by-step checklist for doing just that, alongside instructions for completing each discrete task. If the user wants a step explained in more detail, they can run it through the tool again and have the job broken down into even smaller steps. While there are some kinds of managers who might do this naturally, the Magic Workflow tool allows the average person to do it in just a few clicks. 

Co-founder Judie McCarthy said this seemed like an obvious choice because even large firms seem to lack documented workflows that help people keep track of everything they’re supposed to do. Client Hub itself has many workflow templates, but obviously not for everything.

“We knew this was an area where firms are really struggling. They’d come to us and ask for templates for a lot of work we didn’t have templates for. So now they can create 100 different templates with just a simple description,” she said, saying that while it is a small feature, it is one that adds great power to the product. 

Similarly, practice management software provider Taxaroo released its own ZeroTax.ai solution, a chatbot that answers tax questions. People talking to the bot describe a tax problem they’re having, and the ChatGPT-based AI provides possible solutions. If people want to trust the AI’s answer they can, but if they’re not sure about its advice, they can pay $5 to have the answer vetted by a human accountant. 

In effect, the AI acts as a sort of client intake, according to CEO Mario Costanz, as those who pay to have their answers vetted are connected to human practitioners via Taxaroo’s platform. So, for example, if a user has a question about the estate tax, and then wants further guidance from a human, they might be referred to an estate tax expert via the Taxaroo platform. 

“The vast majority of tax and CPA firms, they’re really good at tax, good at getting referrals, but they’re not good at marketing themselves, so giving them these leads of people at exactly the right place and exactly the right time who are having a sticky tax situation they need answers for now — it’s an opportunity to refer them to the firms that have a deeper knowledge and skill to handle these scenarios,” he said. 

Had Client Hub or Taxaroo decided to build their own AI system from scratch, the projects likely would have cost far more money and taken far more time to develop. By licensing ChatGPT, however, they were able to develop solutions not just cheaply but quickly too. Client Hub’s Magic Workflow took around seven weeks to develop; ZeroTax.ai took a few months. 

While the amount of time a software development project can take varies widely depending on the complexity of the project, broad estimates from several different sources that discuss the topic (including Cleveroad, ClearLaunch, SolTech and Decode) say that, on average, a simple project might take between four and six months, a moderately complex project could take up to nine months, and a very complex project could take several years.

In certain cases, development was helped even further along by ChatGPT itself, as one of its main features is its ability to write code based on plain English prompts. People tell it what they want, and ChatGPT can spit out the required code (though the quality cannot be assured).  

Tax and practice management software provider Canopy’s recently released AI assistant solution uses ChatGPT to draft emails based on keywords and phrases. Instead of composing an entire message about, for example, how an audit is getting out of scope and will need additional fees, the user opens the assistant (a button in the email composition window), inputs keywords like “more work was required,” “out of scope,” “higher price” and “Are you OK?” and the software will then draft the email, which users can then further modify to be, for example, more or less formal in tone. 

Overall, development time was just two and a half months, and ChatGPT was one of the reasons. Colin Childs, director of back-end engineering for Canopy, said that while ChatGPT is not the sole author of the solution, it was a significant contributor to the development process, which greatly boosted productivity. 

“We have seen its potential to impact our employees and improve our processes… ChatGPT is really good at this kind of greenfield-type project, which this ended up mostly being, and it did a fantastic job for laying the groundwork, which a senior developer then goes in and fleshes out to make a high-quality product,” he said. 

And while ChatGPT wasn’t used in the coding of Client Hub’s Magic Workflow solution, McCarthy said that engineers are already using it to code future products and features. 

Canopy CEO Davis Bell pointed out that this is only the beginning when it comes to people incorporating AI into their products. He himself is looking forward to seeing what newer models — perhaps ones made by companies other than OpenAI — are able to do and how they can be used in Canopy’s product offerings as well. He clarified that the company is not building a strategy around ChatGPT, but rather AI overall.

“We’re not married to ChatGPT; they were obviously first to market in a major way, but we’ve been watching other players and we plan on using them as they’re released. For AI generally this is just the very beginning. We think that, over time, you’ll see AI embedded within every single module and feature of Canopy. We believe that accountants who learn to leverage AI-powered software will be very effective and we plan on being the company that will arm accountants with it,” he said.