Hemmingen – September 10, 2026 -- German legal-tech firm Kanzlei Kraftwerk says AI agents can reduce manual workload in tax and law firms by up to 80% for suitable processes, allowing firms to absorb higher case volumes without proportional headcount growth.
Tax assessment reviews consume thousands of hours annually
According to founder and CEO Jan Skrzypczak, a standard tax assessment review takes an average of roughly 15 minutes per case. With firms processing four- or five-digit volumes of incoming assessments per year, this workload accumulates to several thousand hours of manual labor annually.
Digital workers automate full process chains, not single tasks
Kanzlei Kraftwerk builds AI agents and automated workflows integrated with DATEV, Microsoft 365, Outlook and document management systems used by tax advisors, lawyers and auditors. Unlike a chatbot used for a single query, the firm's digital workers link multiple sequential steps -- document recognition, cross-referencing data, drafting client or tax authority correspondence, and initiating follow-up actions -- within one continuous workflow. Human staff are only pulled in at defined decision, review and approval points.
Founder brings two decades of enterprise software and AI experience
Skrzypczak has worked in software, AI and digitalization for more than 20 years, including roughly 12 years in international leadership roles at a major software company. He warns that firms failing to automate repetitive work will eventually face the consequences through staff overload, rising personnel costs, or having to turn down new mandates.
Annual financial statement automation targeted next
Kanzlei Kraftwerk has already automated significant portions of the annual financial statement preparation and review process and is now working to close remaining system interface gaps to extend automation across this more complex workflow.
Process analysis, not software selection, determines automation payoff
The firm's approach starts with mapping how often a task occurs, how much staff time it consumes, and which systems are involved -- before selecting between rule-based automation or AI models based on cost, speed, stability and quality rather than brand recognition. Data protection, information security, governance and applicable regulatory requirements are factored into each deployment.