Appeal after just one prompt: DATEV counteracts shortage of trained tax accountants with AI tools based on Azure

The challenge: Shortage of skilled workers is making it harder for tax accountants to do their job

Time to file those dreaded taxes again! What are the legal requirements? What can I deduct? When is it worth submitting an appeal? These are questions that many wage earners have going around and around in their head at least once a year. And we’re talking about a great many heads indeed: in Germany, for instance, some 46 million people pay private income tax, according to the German Economic Institute. Then there are the 4.1 million companies required to pay trade tax—from neighborhood bakeries to global corporations. That makes 46 million tax declarations that have to be filled out and filed every year. Tax accountants help their clients navigate their way through the ever-changing tax landscape, and their workload increases every year.

“Tax accounting firms always have to be up to speed on current tax law, respond quickly to assessments, and be ready to answer last-minute questions. After all, they’re liable for the advice they provide,” says Dr. Robert Helbig, Director of DATEV Inkubator, Technological Innovation & Research, and Generative AI at DATEV. “At the same time, they’re continuously confronted with new documentation requirements, more regulation, and new processes. We’re on hand to provide state-of-the-art solutions to offset the shortage of skilled staff,” Helbig says.

As an association serving the needs of the tax accounting profession, DATEV eG combines extensive knowledge of tax law with IT expertise. “We’ve already developed a large ecosystem of digital solutions for our customers. To keep pace, for instance with shifts in regulatory requirements for tax accounting firms, we’re always developing new ways of generating added value,” Helbig says. “Our best chance of counteracting the shortage of skilled workers is to keep moving and seizing those promising opportunities.” That’s why Helbig and his team used Azure OpenAI Service as the basis for establishing the DATEV AI Workshop. Through this platform, they provide generative AI-based prototypes that help users work more efficiently.

The solution: Co-creating AI solutions based on Azure OpenAI

Six prototypes are available through the AI Workshop: Social Media Assistant, Einspruchsgenerator (appeal generator), DATEV-GPT, Jobinator, Summarizer, and Frag LEA (ask LEA). One day, these could help tax accounting firms with their work. “It all started with Einspruchsgenerator—the development of which was a crucial first step for the AI Workshop,” Helbig explains. “It autonomously spots discrepancies or appeal-worthy issues in a tax assessment. It then checks these discrepancies against the current legal requirements, formulates the case for the appeal, and submits it.” In the past, all these process steps were done manually and took up a great deal of time—time that the DATEV Einspruchsgenerator will save tax accounting firms in the future.

Jobinator provides DATEV customers with a quick and easy way of writing sector-specific job ads to attract new tax professionals. Social Media Assistant drafts Instagram, Facebook, or LinkedIn posts featuring news, marketing activities, or event invitations. DATEV-GPT allows firms to try out generative AI in a secure environment managed by DATEV. “Most of the solutions that emerge from our AI Workshop are based on Azure OpenAI Service and help tax accounting firms boost their efficiency in all manner of ways,” Helbig says. To ensure that prompts for the AI solutions can be answered as quickly as possible, DATEV developed its own AI connector to direct them to whichever Azure data center can respond fastest at that given moment.

For its development activities at the AI Workshop, DATEV relies on collaboration with customers. “We have to be familiar with tax law processes. It’s essential that we understand how our users work, and that we know whether or not our solutions will provide them with maximum support,” Helbig says. “We create prototypes to meet our customers’ needs, and make them available early on to tax accounting firms for testing. Based on their feedback, we make adjustments over several iterative stages until the solution is ready to become a fixed part of the process.” DATEV’s target corridor to take a prototype from development to go-live is no more than six weeks.

“Relying on Azure for our AI Workshop was absolutely the right decision: we have a quick and easy way of trying out new things, we have scalability, and we have access to important AI resources with specified provisioned throughput units (PTUs) such as ChatGPT,” Helbig says. DATEV is currently testing yet more solutions from the AI Workshop—Summarizer and Frag LEA: “Tax accountants are constantly receiving updated notes or the most recent verdicts. Such documents can be up to 30 pages long—a veritable flood of information—and managing them takes up a great deal of time. We speed up the process.” Now it’s simply a matter of dragging and dropping the long text into Summarizer, which produces a summary that’s just a few pages long. In the future, Frag LEA will make it easier for tax accounting firms to mine the LEXinform database for information: the AI combs through the database’s pool of knowledge to provide answers relating to, say, options for deducting a home office—complete with sources.

DATEV can provide its users with six prototypes in next to no time. They all help combat the shortage of skilled tax accountants by trimming time-consuming workflows down into a handful of prompts.