How to automate accounting working papers with AI

TL;DR

AI can automate the most boring accounting tasks by turning 17 manual working papers into automated workflows, reducing human time. The plan debuts by auditing time spent per task, then building a centralized database and agent system to handle repetitive work, while ensuring human review and legal responsibility remain intact.

Transcript

If you're still using spreadsheets, that's like a dead giveaway that you're not where you need to be. Accountants will literally spend an hour doing this. We just did it in >> Never seen that before. Never experienced it. Extremely efficient. >> This is Josh Thomas. His firm Thomas and Co. is one of the fastest growing accountancy and tax advisory ... Read More

Key Insights

• 17 working papers were identified and prioritized for automation, highlighting where time goes and where AI can help most.
• The goal is not to replace humans but to reduce repetitive data entry and speed up the workflow by coupling AI with human review.
• VAT is chosen as the centerpiece because it is common to all clients and can be scaled across other working papers.
• A centralized database and AI agents are introduced to extract data from client records and automate calculations.
• The plan emphasizes measurable gains, such as reducing an hour plus to minutes for targeted tasks.
• Spreadsheets are deemed dead and the database is described as the new backbone for AI in this workflow.
• Knowledge transfer is critical, as much of the firm’s know-how lives in people’s heads and must be encoded for AI to perform well.
• Regulatory responsibility remains with humans since the AI cannot bear legal accountability or sign off on work.

Questions & Answers

Q: How does the video propose starting an AI automation project in a professional services firm?

The project begins with auditing all 17 working papers to map context, frequency, task duration, and difficulty, then selecting the quickest wins to automate first. This approach prioritizes low effort, high impact tasks such as VAT where 100 percent of clients need it, and gradually scales automation to other papers.

Q: What is the main goal of the AI initiative described in the video?

The main goal is to stop wasting the human’s time on manual data entry and repetitive tasks while enabling the AI to handle data extraction, validation, and routing to the right person for review. Humans still sign off and supervise tasks, but the workload is transformed into an AI-assisted workflow.

Q: Why is VAT chosen as the centerpiece for initial automation?

VAT is chosen because it is universally used by all clients and has a consistent workflow, making it an ideal starting point for building scalable AI processes. Automating VAT provides a reliable proving ground for a broader system that can be extended to the other working papers.

Q: What challenges are highlighted about transferring knowledge to the AI system?

A major challenge is the knowledge stored in the heads of accountants and managers. The team must download this tacit knowledge and embed it into the AI so the system understands how the business operates and how to handle exceptions, rather than just processing raw data.

Q: What is the role of the centralized database in the automation plan?

The centralized database acts as the backbone for the AI system, replacing spreadsheets and providing a structured repository for numbers and their proofs. It enables the AI to access consistent data, perform validations, and scale across multiple working papers with standardized rules.

Q: How does the video describe the legal and ethical constraints of using AI in accounting?

The video notes that AI can do the work but cannot legally take responsibility for it. Humans must sign off on deliverables, and the real consequences in a regulated industry necessitate human oversight and accountability to maintain trust and compliance.

Q: What is the envisioned end state after four days of AI development in the video?

The envisioned end state is a fully built system where the firm’s numbers flow from the accounts through the VAT rules into a tracker that sits inside client channels, with human review for exceptions and final sign-off, effectively creating an asynchronous, AI-assisted mini workforce.

Q: What tools and platforms are used to build and deploy the AI solution?

The project uses Kylon as the platform to onboard AI agents, a database-backed workflow, and integration with accounting data. Additional tools mentioned include AIOS workspace, Railway for hosting, Apify for web scraping, and systems for code review and transcription to support the solution.

Summary & Key Takeaways

• The video shows a UK accounting firm using AI to automate 17 types of working papers, with a focus on VAT, month end, and other common tasks. The process starts with measuring time spent on each task and identifying quick wins. A data-first approach replaces spreadsheets with a centralized database.

• A developer and AI agents are used to map tasks, extract data from client records, and route work to humans for review when needed. The aim is to reduce manual data entry and free up accountants for higher value work, while maintaining regulatory compliance and client service.

• The project highlights the human element, including onboarding AI with context from Josh Thomas and the importance of knowledge transfer from people to the AI system, ensuring the system understands how the business actually works. The outcome demonstrates potential efficiency gains without fully removing human oversight.