Episode 2: Guardrails and Global Reach

Season two continues with an episode that swings from a friendly run through the City of London to a genuine warning about AI governance, then zooms out to the pressures facing the profession worldwide.

It opens on the Standard Chartered Great City Race, where twenty ICAEW colleagues joined more than five thousand runners for the annual five kilometre event through London. The real story wasn't the running, though. Afterwards, Sarah Prescott, who chairs the CAW taskforce for wellbeing, equality, diversity and inclusion, organised a joint social with runners from Chartered Accountants Australia and New Zealand, turning a race into a chance to build relationships across the network. CAW even sponsored a trophy for the fastest team. CA ANZ took it home this year, helped by an individual time of sixteen minutes thirty.

From there, the tone sharpens. ICAEW has published six safeguards every organisation should build into an AI agent deployment, and the underlying message is blunt: most AI failures trace back to the organisation, not the model. The list covers limiting an agent's access to only what it needs, sandboxing before going live, naming a human owner for every agent, keeping backups the agent can't reach, building in a kill switch, and having a clear escalation plan. That last point carries real weight. EU breaches can mean fines of up to thirty five million euros, or seven per cent of global turnover, and recent research found that none of the leading AI models currently meet an acceptable bar for compliance with EU AI and privacy law. This isn't a job to hand off to IT. It's a leadership responsibility.

The episode closes with a global scan from ICAEW's mid-year review, and the picture is a profession under pressure from every direction at once. Tariff uncertainty is now a finance problem in North America. Talent shortages are constraining Chinese firms expanding overseas. Europe's sustainability reporting timeline has slipped again, even as a harmonised EU framework moves toward a possible 2027 launch. Australia is pushing ahead with mandatory climate disclosures, and Southeast Asia is focused on building its next generation of finance professionals.

Three stories, three different scales. But the thread holds: relationships, governance and global awareness are all doing more work than usual right now.

Transcript

SPEAKER_00 0:00

Hello and welcome back to Chartered Accountants Global Update, the podcast from Chartered Accountants Worldwide. This episode is a bit of a mixed bag, a race through the City of London, a warning about AI agents running loose in your systems, and a tour of what's actually happening to the profession, market by market, around the world. The profession is having to move fast, on more fronts than usual. Let's start somewhere lighter. Last week, 20 ICAEW colleagues laced up for the standard chartered Great City Race, a 5-kilometer run through the City of London, alongside more than 5,000 other participants from firms across the capital. It wasn't just a run. Afterwards, Sarah Prescott, who chairs the CAW Task Force for Wellbeing, Equality, Diversity, and Inclusion, put on a joint social with runners from Chartered Accountants Australia and New Zealand, giving people from both institutes a chance to connect properly, away from conference rooms and video calls. CAW sponsored a trophy for the fastest team, based on combined times of each institute's quickest male and female runner. ICAEW's best were Tom Carson and Mia Druitt, both in at 22 minutes 30. But CAANZ took the win, on the back of an individual time of 16 minutes 30. Fast. It's a small story, but it's worth including for a reason. Well-being and cross-institute relationships don't happen by accident. Someone has to organize the social, sponsor the trophy, coordinate the sign-up. That's the kind of quiet groundwork that keeps a global network feeling like a network, and not just a shared logo. Now for something with a bit more edge to it. ICAEW has published a piece on AI agents, and its central claim is blunt. Most AI failures aren't really about the technology. They're about the organization around it. The article sets out six safeguards, and they're worth knowing even if you're not the one configuring the systems. First, limit the blast radius. Give an agent only the access it needs to do its job, nothing more. An agent that summarizes emails has no business holding delete permissions on a database. Second, sandbox it. Before anything goes live, it runs in an isolated environment that mirrors the real system. So if it hallucinates or misfires, nothing real gets touched. Third, name an owner. Every agent needs a person accountable for its access, its performance, its output. Skipping this step is apparently one of the most common ways organizations lose control as they scale up. Fourth, keep backups the agent can't reach. If you've heard the stories about agents wiping inboxes or deleting company data, this one explains itself. Fifth, build in a kill switch. A hard control outside the agent's own reasoning, so it can be cut off instantly if something goes wrong. And sixth, have an escalation plan, including a clear grasp of the legal exposure. Under EU rules, breaches can mean fines up to 35 million euros or 7% of global turnover, whichever is higher. And research from the ITOS Research Foundation this year found that none of the leading AI models currently meet an acceptable standard of compliance with EU AI and privacy law. The point isn't caution for its own sake. It's that AI governance is turning into exactly the kind of discipline accountants already practice in financial reporting and risk management. This is a leadership responsibility, not something to quietly hand to IT. Which brings us to the big picture. This mid-year review pulls together what's happening across the profession region by region, and read together, it's a useful map of where the pressure is building. In North America, shifting U.S. tariff policy has become a finance problem, not just a supply chain one, with firms now having to quantify and disclose the financial impact of trade uncertainty. In China and North Asia, the story is talent. As Chinese firms expand overseas, demand for finance professionals fluent in global standards has outpaced supply, a real constraint on cross-border growth. Europe's sustainability reporting timeline has slipped again, giving many organisations breathing room but extending the uncertainty, while a proposed harmonised EU framework moved through the legislative process and could land as early as 2027. Australia is pushing ahead with mandatory climate disclosures aligned to IFRSS1 and S2, putting it at the front of the pack. And Southeast Asia is focused squarely on building the next generation of internationally capable finance professionals. Different regions, different pressures. But the common thread is unmistakable. Governing technology responsibly, reporting sustainability credibly, and operating across shifting trade rules, all while building the talent pipeline to actually do it. So that's today's episode. A reminder that this profession runs on relationships as much as on rules, that new tools like AI agents demand old disciplines like ownership and accountability, and that wherever you're listening from, the pressures shaping your day job are shaping someone else's on the other side of the planet, too. Thanks for listening, and we'll catch you next time on Chartered Accountants Global Update.