Professional services / Audit

Deloitte Tohmatsu Group: Using in-house generative AI tools and AI agents in consulting and professional work

Company
Deloitte Tohmatsu Group
Country
Japan
Adoption stage
In operation
Source published
Date basis
The date the source was published. It can differ from the date adoption started.
How the source was checked
Read the full source text

The work problem

According to an article Deloitte Tohmatsu Group published on its official website on September 3, 2025, getting the most out of generative AI requires a robust foundation and effective use in work, and an environment that sustains AI transformation is needed alongside tool development. To this end, the group organized its environment around three points: development, talent development and adoption. Previously, consultants had to devise prompts themselves for each work step, such as organizing issues or designing key questions, and enter them directly into generative AI to obtain outputs.

Technology and data

Since 2023 the group has had a special task force reporting directly to the CxOs that continuously tests the latest generative AI technologies and quickly introduces practical ones. It developed four in-house tools, including a chatbot-style generative AI tool and a tool that automatically creates proposal materials, deployed them company-wide, and built RAG for using internal data and a load-balancing mechanism for heavy access. It runs specialist training for engineers, a library of about 100 or more prompts, mandatory e-learning and training for new graduates and mid-career hires, and it formed super-user teams in each organization that redesigned organization-specific work on the premise of using generative AI. The group has operated AI agents with the company-wide rollout completed as of March 2025. The agents consistently and automatically generate prompts suited to each task, use internal materials and review the outputs. The group also introduced an HR agent that automates labor management and HR work while properly managing personal information.

Results

According to the article, as of July 2025 about 12,000 staff across the group were using in-house generative AI tools in their professional work, and the resulting reduction in working hours reached about 100,000 hours a month. The AI agents are already in actual operation across a wide range of work areas, and the group said they have raised the quality of generative AI outputs and greatly shortened working time. The group explained that the prompt library and training have been especially helpful in developing young consultants, and wrote that it aims to lighten the burden of HR work and improve accuracy through the use of data with the HR agent.

Limits and open questions

The article does not explain in its text how the figure of about 100,000 hours a month was calculated, and does not disclose the names of the AI models used or quantify the quality gains and time reductions from the agents.

Sources

Compiled from public sources. These are not results from ATF Works customers.

Read original (opens in a new tab)