Company Interview / Why AI's biggest challenge isn't the technology | the global view

Key points:

AI seen as a net positive, but organisational change lags technology adoption Rising AI costs not yet matched by clear productivity or profit gains Proprietary “fiduciary grade AI” positioned as safer, cheaper alternative to frontier models

Steve Hasker from Thomson Reuters sets out a cautiously optimistic view on artificial intelligence, arguing it is a net help for knowledge workers in law, tax, accounting and audit. Hasker states that tools such as ChatGPT and Claude are already widely used, particularly in Australia, but contends that organisational change management lags adoption by up to 18 months. In his view, this delay explains why CFOs see rising AI token costs without equivalent gains in speed or efficiency.

Hasker highlights a widening gap between AI spend and bottom-line impact, suggesting some high-profile job cuts at large technology companies may be “AI washing” rather than genuine productivity gains. He argues that meaningful benefits will only emerge once workflows, handoffs and human–machine roles are systematically redesigned.

Hasker says Thomson Reuters has built its own large language model for the legal profession, “Thomson”, based on proprietary datasets including Westlaw, Practical Law and the Reuters news file. He claims this model outperforms frontier models on complex legal tasks and operates at materially lower cost, due to owned, curated content. Hasker promotes the idea of “fiduciary grade AI” for professions where accuracy and liability are critical, noting rapid adoption of Westlaw Advantage and the revamped CoCounsel legal assistant in the US, UK and now Australia