The Australian arm of consultancy firm KPMG wrote a 100-page prompt to create an agentic system that prepares tax advice far faster than humans.
Speaking at analyst outfit Forrester’s APAC Technology & Innovation Summit on Tuesday, KPMG chief digital officer John Munnelly said the firm’s “life changed” when ChatGPT debuted in late 2022, because he quickly understood it was a tool the consultancy could not ignore.
However early experiments produced “really scary” results including the discovery of a single document on KPMG servers that listed thousands of employees’ credit card numbers.
“That absolutely scared the pants off me,” he said. KPMG therefore stopped its experiments and blocked ChatGPT while it assessed the risks AI posed.
It does what our team used to do in about two weeks, in a day.
Not many weeks later, on his return from a holiday, Munnelly turned on his phone to find a long list of missed calls: A graduate staffer had taken a screenshot of the KPMG IT environment that showed it had blocked ChatGPT and shared it on social media with a derisive message about the firm’s attitude to innovation. A prominent business newspaper ran a story about the post.
Happily, KPMG was already negotiating new software licenses with Microsoft, which offered access to OpenAI’s tools.
The firm started building a private AI platform, and apps that used it. So did other branches of KPMG around the world, sometimes unwittingly replicating work done by their peers in other nations. The firm eventually decided on a combined approach and created a tool called “KPMG Workbench” that offers retrieval-augmented generation (RAG), LLMs, and agent hosting to all member firms around the world.
KPMG decided it was wise not to assume that any single vendor would dominate LLMs, so Workbench uses models from OpenAI, Microsoft, Google, Anthropic, and Meta.
Through 2023, the firm focused on training staff on how to use chatbots and write effective prompts.
In 2024, it started building agents, including the TaxBot mentioned above.
Munnelly said building that bot started with locating tax advice written by partners, which he said was “stored all over the place” – often on tax partners’ laptops. KPMG found as much of that advice as it could and placed it in a RAG model along with Australia’s tax code to produce an Agent that creates tax advice.
“It is very efficient,” Munnelly told the Forrester conference. “It does what our team used to do in about two weeks, in a day. It will strip through our documents and the legislation and produce a 25-page document for a client as a first draft.
“That speed is important,” he added. “If we have a client who is about to do a merger, and they want to understand the tax implications, getting that knowledge in a day is much more important than getting it in two weeks’ time.”
“That is really changing our business and how we work.”
Munnelly said KPMG built the agent by writing a 100-page prompt it fed into Workbench. The Register asked for details of the prompt and Munnelly said a substantial team worked on it for months, and the resulting agent asks for four or five inputs before it starts working on tax advice, then asks a human for direction before generating a document.
Only tax agents can use the tool, because its output is not suitable for people without deep tax expertise.
Munnelly thinks 100-page prompts probably won’t be necessary in future, because KPMG has since built an agent runtime service that allows agents – writers, editors, and managers – to interact on tasks such as conducting research or writing “credentials” – KPMG-speak for project summaries it can re-use on other jobs.
The chief digital officer said KPMG has deployed agents that do frustrating and time-consuming work people would rather avoid, and that staff surveys suggest employee satisfaction has risen as AI frees them to spend more time working on challenging tasks, leading them to rate the firm as more innovative.
“They just don’t want to do the boring stuff,” Munnelly said. “They want to get out there and help clients with chewy problems.”
Agents have also produced new revenue, because clients asked to buy them. “We have additional revenue streams that we didn’t expect,” Munnelly said, adding that KPMG sees benefits in time, quality, and revenue from its use of AI.
He admitted that it’s hard to measure the benefits of AI use at the firm, but didn’t mention job losses and suggested KPMG will enthusiastically explore new ways to use AI. ®