Weekly AI Wrap: Cisco Cuts 4,000 While Posting Record Revenue, GPT-5.5 Lands in M365, and Regulators Crack Down on AI-Washing
A big week in AI — and not just because of the federal budget. While Canberra was reshaping NDIS and aged care funding, the global AI industry was having its own structural moment: the single biggest corporate AI capex number in history, a cybersecurity initiative from Anthropic that's rattling regulators, a new C-suite role that's gone from novelty to near-universal in twelve months, and Australia's consumer watchdog adding AI-washing to its enforcement priorities. Here's the week's read.
Cisco Cuts 4,000 Jobs the Day It Reports Record Revenue — and Its Stock Jumps 16%
Cisco announced the elimination of nearly 4,000 roles — under 5% of its global workforce — on 14 May, the same day it reported its highest-ever quarterly revenue of $15.8 billion, up 12% year-on-year. The restructuring is expected to cost up to $1 billion in pre-tax charges. The market reaction: shares rose more than 16% in after-hours trading.
The reason investors cheered a layoff announcement alongside record revenue is the AI order pipeline. Cisco has secured $5.3 billion in AI infrastructure orders from hyperscalers so far this fiscal year — and has now raised its full-year AI order expectation to $9 billion, up from $5 billion previously. CEO Chuck Robbins was direct: "The companies that will win in the AI era will be those with focus, urgency, and the discipline to continuously shift investment toward the areas where demand and long-term value creation are strongest." The roles being eliminated are in traditional networking and general operations. The capital is being redirected to silicon, optics, security, and AI infrastructure — the hardware layer underneath the AI boom.
The Cisco story is a cleaner version of what's happening across the sector: profitable, growing companies treating headcount as a capital allocation decision rather than a growth signal. Across tech broadly, more than 92,000 workers have been laid off in 2026, with AI restructuring cited as the primary driver for the first time.
GPT-5.5 Instant Is Now Inside Microsoft 365 — What It Actually Means for Finance Teams
Microsoft rolled out GPT-5.5 Instant to Microsoft 365 Copilot this week — available across Word, Excel, Outlook, and Teams for licensed users. The model is optimised for speed and everyday task quality: faster document summarisation, cleaner spreadsheet analysis, more accurate drafting with less back-and-forth. It's accessible in the Copilot Chat model selector as "GPT-5.5 Quick response" and is also rolling out in Copilot Studio for organisations building internal AI agents.
For finance teams already inside the Microsoft 365 ecosystem, this is a meaningful practical upgrade — particularly for Excel-based analysis and Outlook-driven workflows. The more significant shift underneath the model update is the agent architecture: GPT-5.5 Instant in Copilot Studio means finance teams can now build faster, more reliable internal agents for tasks like management report drafting, variance flagging, and document processing — without touching external AI tools or APIs.
Anthropic's Claude Mythos Preview Is Finding Zero-Day Vulnerabilities Nobody Knew Existed
Anthropic's Project Glasswing — a controlled security initiative — gave a small group of major technology and financial institutions access to Claude Mythos Preview, an unreleased frontier model, specifically to identify critical software vulnerabilities before malicious actors could exploit them. During testing, the model reportedly discovered thousands of zero-day vulnerabilities, including a 27-year-old bug in OpenBSD that had gone undetected since 1999.
The institutions involved include AWS, Apple, Cisco, Google, JPMorgan Chase, and Microsoft. The controlled release structure reflects the dual nature of the capability: the same model that can identify vulnerabilities in defensive security work can, in the wrong hands, be used to find and exploit them offensively. ASIC flagged exactly this risk in its 8 May letter to the financial sector — frontier AI capabilities materialising "incredibly quickly" in ways that outpace existing defensive infrastructure.
The Glasswing announcement strongly suggests that the most advanced AI systems are already operating several capability levels beyond what's accessible through consumer products or standard APIs. The gap between what's publicly available and what's running in controlled enterprise environments is wider than most people assume.
The Chief AI Officer Is Now Standard Issue — 76% of Organisations Have One
IBM's 2026 survey found that 76% of organisations now have a Chief AI Officer — up from just 26% last year. That's not a gradual trend. That's a structural shift in how organisations are treating AI governance responsibility, compressed into twelve months. The CAIO role is distinct from a CTO or CDO: it's specifically focused on AI strategy, adoption governance, risk management, and organisational change around AI — not on building models directly.
The Australian Public Service formalised this direction in its National AI Plan, which requires a Chief AI Officer across every agency. The logic is the same as the corporate version: AI adoption without named accountability at the executive level creates governance gaps that accumulate quietly and surface expensively. The APRA review this week identified exactly those gaps in regulated financial services entities — governance frameworks lagging behind the pace of adoption.
For mid-market organisations that can't justify a standalone CAIO, the practical equivalent is assigning AI governance responsibility explicitly to an existing executive — typically the CFO or COO — rather than leaving it distributed across teams with no single owner. The role doesn't require deep technical expertise. It requires understanding what decisions AI is informing, what data it's touching, and what happens when it's wrong.
ACCC Has Made AI-Washing an Enforcement Priority for 2026-27
The Australian Competition and Consumer Commission has included AI-washing in its 2026-27 compliance agenda — specifically targeting organisations that make misleading claims about the AI capabilities of their software. The enforcement focus is on companies describing basic automation or rules-based logic as machine learning, or overstating the sophistication of AI-driven features in ways that mislead customers or procurement decisions.
The ACCC's position is that AI-washing constitutes deceptive conduct under the Australian Consumer Law — and the penalties for deceptive conduct are significant. For organisations purchasing AI-enabled software or services, this adds a layer of due diligence: understanding what the product actually does versus what the vendor claims it does is now a procurement risk question, not just a technical one.
This matters particularly for finance teams evaluating AI tools for management reporting, FP&A, or payroll automation. The market is saturated with "AI-powered" claims that range from genuine machine learning to rebranded Excel. Knowing the difference — and being able to document the basis for your procurement decision — is increasingly important.
Using AI in Your Finance Function and Want to Get the Governance Right?
PFL works with Australian NFP, NDIS, and SME finance teams on AI automation — including the governance frameworks that make adoption sustainable and defensible. If your team is using AI tools and the question of who owns the risk hasn't been answered yet, let's talk.
Get in Touch with PFL →
Comments
Post a Comment