AI in accounting: e-invoicing and automation

Since January 1, 2025, all domestic B2B companies in Germany must be able to receive e-invoices. At the same time, 57 percent of all tax advisors are over 50 years old, and 71 percent of tax offices report acute staffing shortages. Two developments that would each be far-reaching on their own—together they create a transformation pressure that no finance department can ignore. The good news: artificial intelligence is delivering the tools right now to not only withstand this pressure but to turn it into a genuine competitive advantage. This article shows what AI can already accomplish in accounting today, which tools facilitate getting started, and how the role of the accountant is fundamentally changing.

E-invoicing mandate 2025/2026—what is changing and why AI is the key

The e-invoicing mandate is no longer a distant future scenario. Since the start of 2025, the rule is: every company with B2B business in Germany must be able to receive structured e-invoices in XRechnung or ZUGFeRD format. From 2027, larger companies must also actively send in structured format, and from 2028, the sending mandate applies to all.

What initially looks like a purely technical conversion is in reality a fundamental process transformation. Because a true e-invoice is not a PDF file in an email attachment. It is a machine-readable XML document that can be processed directly by accounting systems—without manual data entry, without media breaks, without typos.

This is precisely where AI becomes the key. Because the e-invoicing mandate provides, for the first time across the board, structured data in a uniform format. That is exactly the prerequisite that AI systems need to deploy their full power. According to Kay Dietrich, manually processing a single invoice costs 5 to 15 minutes. With 200 incoming invoices per month, that adds up to 50 working hours—more than a full working week spent solely on repetitive tasks.

The e-invoicing mandate is therefore not just a regulatory requirement. It is the starting signal for an entirely new way of thinking about and organizing accounting.

What AI can already do in accounting today

The days when AI in accounting was purely a future promise are over. Three core areas are already measurably benefiting from intelligent automation.

Receipt capture: from manual entry to automatic extraction

Receipt capture is the area where AI delivers the greatest immediate impact. Modern systems combine OCR technology (Optical Character Recognition) with semantic AI that does not just recognize individual characters but understands context. The result: automation rates of 95 to 99 percent in data extraction from incoming invoices.

The workflow is remarkably lean. A system like livestep.ai describes the typical process in five steps: select invoice, check PDF, automatically extract data, create posting suggestion, review and confirm. The AI handles the routine; the human only reviews. An integrity check ensures that the extracted data is consistent and plausible.

For companies, this means: instead of 10 minutes per invoice, 30 seconds for approval suffice. With 200 invoices monthly, that saves approximately 30 to 45 hours—every single month.

Account assignment: learning systems instead of rigid rules

Traditional accounting software works with fixed assignment rules: Supplier X is always posted to Account Y. This works for 80 percent of cases. The remaining 20 percent—special cases, new suppliers, unusual invoice items—land on the accountant’s desk.

AI-powered account assignment goes a step further. Learning systems analyze historical posting patterns, recognize relationships between invoice contents and accounts, and generate automatic posting suggestions. The longer the system is in use, the more precise the suggestions become. Haufe Finance reports that modern systems already achieve accuracy rates above 90 percent for automatic account assignment.

The decisive advantage over rule-based systems: the AI learns from corrections. When the accountant changes a suggestion, the system remembers the correction and applies it automatically in the future. The system gets better with every posting.

Reporting: from half-day reports to real-time dashboards

Reports that previously required half a day of manual work are done in minutes with AI support. But the real advancement lies not in speed but in depth. AI-powered reporting systems can detect anomalies, identify trends, and create forecasts that a manually created report would never deliver.

An example: instead of creating a liquidity report once a month, an AI dashboard delivers a running cash flow forecast that automatically balances incoming payments, outstanding receivables, and planned expenditures. This is not science fiction—tools like Candis and Yokoy already offer this functionality today.

Haufe Finance describes this development as the path to “Continuous Closing”: instead of deadline stress at month-end or year-end, closings are continuously prepared because data is permanently current and verified.

Tool comparison: AI solutions for automated accounting

The market for AI-powered accounting tools is growing rapidly. The following table provides an overview of four established solutions that are particularly relevant for the German market.

• Tool: GetMyInvoices | Target Group: Midsize companies, freelancers | Strengths: OCR recognition, automatic receipt retrieval from over 10,000 portals, easy setup | DATEV Integration: Yes, bidirectional | Price (from): approx. €15/month
• Tool: Candis | Target Group: Midsize companies (5–250 employees) | Strengths: Automatic AI-powered account assignment, approval workflows, learning posting suggestions | DATEV Integration: Yes, real-time sync | Price (from): approx. €49/month
• Tool: DocuWare | Target Group: Midsize companies | Strengths: Workflow automation, DMS integration, audit-proof archiving, GoBD-compliant | DATEV Integration: Yes, via interface | Price (from): On request
• Tool: Yokoy | Target Group: Midsize companies, enterprises | Strengths: Expense management, travel expense processing, automatic compliance verification | DATEV Integration: Yes, via API | Price (from): On request

Important when choosing: All four tools offer DATEV integration, which is crucial for the German market. The difference lies primarily in the target group and feature scope. GetMyInvoices is excellent as an entry point for small businesses and freelancers. Candis offers the best mix of AI account assignment and user-friendliness for midsize companies. DocuWare is the choice for companies that need a comprehensive document management solution. Yokoy covers the entire expense area and is thus particularly interesting for companies with high travel and expense volumes.

Additionally, Kay Dietrich describes the use of MCP servers (Model Context Protocol) that connect AI systems directly to DATEV. This technology enables an AI to not only read invoices but actively suggest postings in DATEV and, after approval, execute them as well. This is an approach that is becoming particularly relevant for tax firms and larger accounting departments.

The future: from accountant to data steward

Perhaps the most important insight from current developments: AI does not replace the accountant. It fundamentally changes their role. Haufe Finance and Karsten Mitzinnek describe a transformation that will be complete by 2030 and will create new professional profiles.

• Role Today: Accountant | Role by 2030: Data Steward | Core Activity: Ensuring data quality, training AI systems, handling exceptions
• Role Today: Controller | Role by 2030: Business Partner | Core Activity: Strategic analyses, data-driven decision support for management
• Role Today: Accounting Clerk | Role by 2030: Automation Process Manager | Core Activity: Configuring workflows, maintaining automation rules, monitoring system integration
• Role Today: Tax Advisor | Role by 2030: Strategic Tax Architect | Core Activity: Tax optimization, AI-powered scenario planning, compliance monitoring

This transformation is not a threat but a necessity. Demographic change is creating gaps that simply cannot be closed without technology. When 57 percent of tax advisors are over 50 years old and 71 percent of tax firms report staffing shortages, the question is not whether but how quickly automation advances.

The critical point: the new roles are more demanding and more value-creating than the old ones. A Data Steward who ensures the quality of financial data and trains AI models is more valuable to a company than an accountant who manually enters receipts. A Business Partner who delivers data-driven decision-making foundations to the CEO contributes more to value creation than a controller who compiles monthly reports.

The FAZ reports that 15 AI startups are already working on automation solutions in the tax advisory space alone. This shows: the market is responding to demographic pressure with innovation. Companies that position themselves now will secure the best talent and the most efficient processes.

Practical example: how a midsize tax firm saves 50 hours per month

To make the theory tangible, here is a concrete example. A tax firm in southern Germany with 12 employees serves approximately 180 clients. About 200 incoming invoices per month need to be captured, assigned to accounts, and posted. Additionally, there are 60 to 80 outgoing invoices and ongoing client correspondence.

Starting situation before automation:

• Receipt capture: average 8 minutes per invoice (manual entry, assignment, filing)
• Account assignment: average 3 minutes per receipt (look up account, assign cost center)
• Monthly reports for clients: 45 minutes each for manual compilation
• Total effort per month: approximately 85 hours for these three activities

After introducing an AI-powered accounting solution with DATEV integration:

• Receipt capture: automatic extraction in under 30 seconds, only approval by employee (1 minute per receipt)
• Account assignment: AI posting suggestions with 92 percent accuracy, only exceptions handled manually (average 20 seconds per receipt)
• Monthly reports: automatically generated, only plausibility check (10 minutes per report)
• Total effort per month: approximately 35 hours

Result:

• Metric: Time effort per month | Before: 85 hours | After: 35 hours | Savings: 50 hours (59 percent)
• Metric: Cost per invoice (internal) | Before: approx. €6.50 | After: approx. €1.80 | Savings: 72 percent
• Metric: Error rate in account assignment | Before: approx. 4 percent | After: under 1 percent | Savings: 75 percent
• Metric: Time to monthly close | Before: 8 working days | After: 3 working days | Savings: 63 percent

The 50 hours saved per month are now used by the firm for higher-value activities: individual tax advisory, client consultations, and strategic tax planning. Three employees have undergone internal training and now take on the role of Data Steward—they maintain the AI models, review exception cases, and ensure the data quality that keeps the entire system reliable.

Frequently asked questions

Is my company too small for AI in accounting?

No. For small businesses and freelancers in particular, getting started pays off because the relative time savings are greatest. A freelancer who processes 30 invoices per month and saves 10 minutes per invoice gains back 5 hours—time that can flow directly into revenue-generating activities. Tools like GetMyInvoices start at just €15 per month and can be set up in under an hour.

Do I need programming skills for AI-powered accounting?

No. All tools mentioned in this article work with graphical interfaces and require no programming skills. Integration with DATEV also runs via pre-configured interfaces with most providers. For those who want to build custom workflows beyond that, no-code platforms like n8n can automate accounting processes via drag-and-drop.

How secure are my financial data with AI tools?

Data privacy and data security are central issues. Watch for the following criteria: GDPR compliance, server location in Germany or the EU, encryption during transmission and storage, and a clear policy on whether and how the provider uses your data for AI training. All tools mentioned in this article meet these criteria. Those who need maximum control should prefer on-premise solutions or German cloud providers.

Does AI replace my tax advisor?

No, but it fundamentally changes the collaboration. AI takes over the repetitive tasks—receipt capture, account assignment, standard reports. The tax advisor can then focus on what they are trained for: individual advisory, tax optimization, and strategic planning. The FAZ reports that AI startups aim not to replace tax advisory but to expand it. The tax advisor of the future works with AI, not against it.

What does it cost to get started with AI-powered accounting?

Getting started is more affordable than many expect. Simple receipt recognition tools start at €15 to €50 per month. More comprehensive solutions with automatic account assignment and workflow automation range from €50 to €200 per month. Enterprise solutions for larger companies are individually priced. What matters is the ROI: with time savings of 50 hours per month and an internal hourly rate of €40, even a solution costing €200 per month pays for itself within just a few weeks.

References

• Haufe Finance / Karsten Mitzinnek (March 2026): “KI definiert Rollen im Finanzbereich neu” [in German]—Analysis of the transformation of accountant and controller roles through 2030, Continuous Closing and Semantic AI in receipt capture. https://www.haufe.de/finance
• Kay Dietrich Blog (March 2026): “KI-Automatisierung für den Mittelstand—5 Prozesse die sofort profitieren” [in German]—Practice report on invoice processing with MCP server and DATEV integration, including time savings calculation. https://www.kaydietrich.com/blog
• Frankfurter Allgemeine Zeitung (FAZ) (March 2026): “15 KI-Start-ups wollen die Steuerberatung automatisieren” [in German]—Analysis of demographic change in the tax industry and the role of AI startups. https://www.faz.net
• Haufe (March 2026): “KI in Finanzabteilungen: Vergleich der Tools für Rechnungsverarbeitung” [in German]—Market overview of GetMyInvoices, Candis, DocuWare, and Yokoy with DATEV integration assessment. https://www.haufe.de/finance
• livestep.ai (March 2026): “Rechnungen automatisch buchen mit KI” [in German]—Step-by-step guide to AI-powered receipt processing with integrity verification. https://www.livestep.ai