KPMG WRITES A 100-PAGE PROMPT TO CREATE AN AI TAX – TAXBOT NOW DOES 2 WEEKS’ WORK IN JUST 1 DAY

• KPMG has successfully announced the development of TaxBot, an AI agent specialized in tax consulting, built with a 100-page long prompt after months of work.
• The project originated from the need to digitize a large volume of tax advice written by company partners, which was scattered across internal laptops.
• Following an incident where sensitive data was found unprotected on servers, KPMG temporarily suspended AI experiments and built its safe, proprietary platform called KPMG Workbench.
• Workbench integrates retrieval-augmented generation (RAG) and multiple large language models (LLMs) from providers such as OpenAI, Microsoft, Google, Anthropic, and Meta, avoiding reliance on a single entity.
• TaxBot was developed at KPMG Australia in 2024, operating in a closed environment, using real consulting data and Australian tax laws.
• This bot can request user input, consult with experts, and generate a 25-page draft for clients in just one day – whereas previously it took two weeks.
• Only licensed tax professionals are permitted to use TaxBot, as its output is not suitable for general use without professional supervision.
• Benefits include: increased work efficiency, reduced repetitive tasks, enhanced employee satisfaction, and creating additional revenue from selling similar agents to clients.
• However, KPMG admits it is difficult to accurately measure efficiency and is still evaluating long-term impacts.

Here’s the English translation of your text:

📌 KPMG’s TaxBot is a prime example of AI application in the field of taxation, built with prompts up to 100 pages long and capable of generating a 25-page advisory report in just one day. By integrating RAG and multi-vendor LLMs, the system delivers high efficiency, reducing turnaround time from 14 days to only 1. Access to TaxBot is restricted to licensed tax practitioners, as its outputs are not suitable for general use without professional oversight. However, KPMG acknowledges that measuring effectiveness precisely is difficult and continues to assess its long-term impact.

Tổng hợp.