KPMG’s TaxBot development consumed months of drafting a 100-page promptPartner-written tax advice had been scattered across countless laptopsKPMG Workbench hosts multiple LLM models from competing vendors<\/p>\n
When large language models began attracting global attention in late 2022, KPMG\u2019s digital leaders immediately recognized potential benefits but also major risks.<\/p>\n
Chief digital officer John Munnelly admitted that first experiments with ChatGPT produced \u201creally scary\u201d results, including the discovery of sensitive financial data sitting unsecured on internal servers.<\/p>\n
<\/p>\n That incident caused the firm to suspend experiments, restrict access to public AI tools<\/a>, and reassess the dangers that uncontrolled deployment might introduce.<\/p>\n You may like<\/p>\n Building a private AI platform<\/p>\n KPMG subsequently began constructing a closed environment for AI work, supported by software licenses that allowed access to OpenAI and Microsoft<\/a> systems.<\/p>\n This move gave the consultancy a chance to design applications within safer boundaries, eventually leading to a platform called KPMG Workbench.<\/p>\n The system combined retrieval-augmented generation, multiple LLM<\/a> options, and agent hosting capabilities.<\/p>\n Rather than depending on a single vendor, the firm deliberately spread usage across OpenAI, Google, Microsoft, Anthropic, and Meta.<\/p>\n Throughout 2023, extensive effort was devoted to training employees on how to write prompts effectively and interact with AI writer<\/a> systems.<\/p>\n By 2024, the Australian arm<\/a> of KPMG initiated projects to build specialized agents. Among them was the so-called TaxBot, a tool designed to prepare tax advice.<\/p>\n Munnelly explained that development began by gathering partner-written advice that had been \u201cstored all over the place,\u201d often scattered on laptops.<\/p>\n That information, combined with Australia\u2019s tax code, was placed into a RAG model to produce automated drafts. TaxBot, however, was not trivial to construct.<\/p>\n According to Munnelly, its creation required a 100-page prompt, drafted over months by a dedicated team, and ultimately fed into Workbench.<\/p>\n The result is a system that requests several inputs, seeks guidance from human experts, and then generates a 25-page document for client review.<\/p>\n Munnelly claimed the agent now performs tasks that once took two weeks in a single day, a change he described as \u201cvery efficient.\u201d<\/p>\n He suggested that quick turnaround is particularly important for clients engaged in time-sensitive deals such as mergers.<\/p>\n Yet he also emphasized that only licensed tax agents are permitted to operate the tool, acknowledging that output without professional oversight is not suitable for general users.<\/p>\n Beyond efficiency, KPMG argues that the introduction of agents has boosted staff satisfaction, since repetitive tasks can be avoided.<\/p>\n Additionally, some clients have expressed interest in acquiring similar agents, generating revenue streams KPMG did not originally anticipate. Nevertheless, the firm concedes that measuring precise benefits remains difficult.<\/p>\n