Content
summary Summary

A six-month evaluation of Microsoft 365 Copilot in the Australian public sector has shown that the AI assistant can save employees up to an hour per day, though its implementation faces significant hurdles.

Ad

The Australian Digital Transformation Agency (DTA) conducted the trial across 56 government agencies, issuing nearly 6,000 licenses and gathering feedback from more than 2,000 participants. Responses to individual questions were sometimes only in the triple digits, which seems to be an overall weakness of the study.

Despite the generally positive attitude toward the technology expressed by 77 percent of participants, only one-third of users engaged with the AI assistant daily. More than 80 percent (n=827) said they would like to continue using Microsoft 365 Copilot after the trial, which is good news for Microsoft.

Table displaying Copilot for Microsoft 365 usage frequency, based on 811 survey participants.
The majority of users (46%) utilize Copilot multiple times per week, while a quarter use the software daily. Only 1% of participants report not using Copilot at all. | Image: dta.gov.au

69 percent of employees reported getting work done faster, and only 7 percent reported losing time due to fact-checking and content review. The trial found that Copilot worked best for specific tasks like summarizing information, creating initial document drafts, and information retrieval. Microsoft CEO Satya Nadella recently described summarization as "a big deal" for generative AI.

Ad
Ad
Table showing average daily time savings through Copilot for various office activities, based on 330 survey respondents.
Copilot's time savings are highest for summarizing information at 1.1 hours per day, followed by document creation at one hour per day. | Image: dta.gov.au

Middle management and IT staff saw the greatest productivity gains, with around 40 percent of users reporting that they used the time they saved for higher-value work, such as management tasks and strategic planning. 61% of participants agreed that Copilot improved the quality of their work.

Implementation is challenging

The DTA identified several major obstacles during the trial. Technical integration issues, insufficient training in prompt engineering, and unclear legal responsibilities hampered adoption.

The agency also noted worries about data security, information management, and potential bias in AI outputs. 61 percent of managers couldn't reliably distinguish between AI-generated and human-created content.

Some participants expressed concern about the loss of basic skills, such as summarizing text, and the environmental impact of using AI.

Stacked bar chart: User expectations for Copilot across different Microsoft products from 1,141 respondents.
User satisfaction with Copilot varies across Microsoft products. Teams and Word show the highest approval ratings, while Excel receives more moderate evaluations from users. | Image: dta.gov.au

The evaluation highlighted potential workplace equity issues, noting that increased AI adoption could disproportionately affect administrative positions - roles that traditionally serve as entry points for women and marginalized groups.

Recommendation

Based on these findings, the DTA recommends that government agencies develop comprehensive implementation plans that include specific training programs, clear usage guidelines, and active risk management strategies. The agency also suggests that departments analyze their workflows to identify additional beneficial applications for the technology.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • The Australian government tested Microsoft 365 Copilot for six months across 56 agencies with 5,765 licenses. 2,000 people participated in the evaluation.
  • According to surveys and interviews, 69 percent of users found their work was faster, especially when summarizing, creating documents, and searching for information. Seven percent said they needed more time for their work as a result of AI.
  • Despite the efficiency gains of up to an hour per day, there were barriers to adoption: technical integration issues, lack of training, legal uncertainties, and concerns about data security, bias, and environmental impact.
Online journalist Matthias is the co-founder and publisher of THE DECODER. He believes that artificial intelligence will fundamentally change the relationship between humans and computers.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.