Answer-First Lead
Anthropic wants your local bakery to use Claude for payroll, and to prove it’s serious about democratising AI, it just reinstated OpenClaw access too. Meanwhile, CNBC reports that virtually every Fortune 500 company is now tracking how much AI their employees use — logging prompts, tasks, and even “AI work units” like a new productivity currency. In New Zealand, Deloitte has a blunt warning for local firms: your business model needs an AI redesign, and you’re already late. The technology-people relationship is entering a new phase: not just about who uses AI, but who’s watching who use AI.
🔍 THE BOTTOM LINE
Three shifts are happening simultaneously: AI is becoming accessible to the smallest businesses (Anthropic’s play), AI is becoming trackable by the largest employers (the surveillance play), and AI is becoming legally regulated at the state level (Colorado’s law). The common thread? Technology is no longer a choice — it’s something that happens to you, and you need to decide whether you’re driving or being driven. New Zealand’s business community, per Deloitte, is still trying to find the steering wheel.
📰 Stories
1. 🏪 Anthropic Launches Claude for Small Business: Plug-and-Play Automation for the Main Street
The story: Anthropic launched “Claude for Small Business” — a package of connectors and pre-built workflows designed for small business owners who don’t have dedicated IT teams. The offering integrates Claude Cowork (Anthropic’s task-executing chatbot launched in January) with Intuit QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365. It comes with 15 pre-built skills and an equal number of plugins covering payroll planning, cash flow forecasting, month-end close, customer follow-ups, and marketing campaigns.
A retailer, for example, could set up an automation that checks QuickBooks revenue against HubSpot ad performance and launches a marketing campaign when sales slow — all through conversational prompts.
Xero also launched a live integration with Claude, connecting the accounting platform’s financial data directly.
Why it matters: Big companies have had AI consultants and integration teams for two years. Small businesses — the backbone of every economy — are only now getting tools that don’t require a computer science degree to deploy. This is where the real productivity gains will come from, not from Fortune 500 enterprises adding another AI chatbot to their stack.
Sources: SiliconANGLE, Anthropic Blog, The Register, PYMNTS, The Deep View, 9to5Mac, The Paypers
2. 🔄 Anthropic Reinstates OpenClaw and Third-Party Agent Usage — With Agent SDK Credits
The story: After a controversial April decision to ban third-party agent tools (including OpenClaw) from accessing Claude via consumer subscriptions, Anthropic has reversed course. Starting June 15, Pro and Max subscribers will get monthly Agent SDK credits ($20–$200 depending on plan) to use with third-party agent harnesses.
The reversal came after significant community backlash — OpenClaw’s creator was temporarily banned from Claude, sparking debate about the boundaries between consumer and API usage. Anthropic’s new policy essentially formalises a middle ground: subscription users get a dedicated agent usage allowance, separate from chat query limits.
Why it matters: This is Anthropic admitting that the agent ecosystem matters more than trying to capture every use case internally. The community battle over OpenClaw was a test case for how AI companies treat the open-source ecosystem around their models. Anthropic chose collaboration over control — at least for now.
Sources: VentureBeat, TechCrunch, Awesome Agents, Phemex
3. 👁️ Almost Every Fortune 500 Is Tracking AI Usage at Work
The story: CNBC reports that virtually every Fortune 500 company has deployed some form of AI activity tracking for employees. Salesforce alone is logging 2.4 billion “AI work units” — a new metric measuring AI prompts, tasks, and compute consumption per employee. Microsoft’s Viva platform and ActivTrak have added similar monitoring features.
The tracking goes beyond simple productivity metrics. Companies are analysing which employees use AI effectively and which don’t, creating what privacy advocates call a “two-tier workplace” — AI-adopters get more autonomy, AI-resisters get more scrutiny.
Meanwhile, Michigan has introduced a bill (HB 4833) specifically targeting AI employee surveillance, and California’s AB-1883 would require employers to disclose all workplace surveillance tools. The laws are struggling to keep pace.
Why it matters: The surveillance story isn’t just about keystroke logging anymore. It’s about a new form of workplace visibility where every AI interaction becomes a data point in performance reviews, promotion decisions, and even layoff targeting. The phrase “AI productivity” sounds positive until you realise it’s a metric used against you.
Sources: CNBC, TechRepublic, Observer, ACS Information Age, ActivTrak, California Legislature
4. 🇳🇿 Deloitte Warns NZ Firms: Redesign for the AI Era — You’re Already Late
The story: Deloitte has published a new report warning New Zealand businesses that the window for competitive advantage through AI adoption is narrowing rapidly. The analysis argues that incremental improvements — adding ChatGPT to existing workflows — aren’t enough. Deloitte says NZ firms need fundamental business model redesigns, including new revenue models, cost structures, and talent strategies specifically built around AI capabilities.
The report comes as BusinessDesk reported that the “AI land rush” is accelerating across NZ, with Zuru, One NZ, and the All Blacks all featuring AI use cases at AWS’s recent conference. But Deloitte’s warning suggests most firms are still in the experimentation phase while their international competitors are already in production.
Why it matters: New Zealand’s economy runs on small-to-medium businesses that are typically 2-3 years behind global tech trends. Deloitte is essentially saying that gap is now existential, not just inconvenient. The country can’t afford to treat AI adoption as something to “get around to” — especially when our largest trading partners (China, Australia, US) are embedding AI into policy and trade.
Sources: Deloitte Insights (via RNZ), BusinessDesk, RNZ
5. ⚖️ Colorado AI Law Finalised — A Watered-Down Version of What Was Proposed
The story: Colorado’s AI law, which began as one of the most ambitious state-level AI regulations in the US, has been finalised in a significantly watered-down version. The original bill would have required impact assessments for any “high-risk” AI system — but intense lobbying from tech companies, insurance firms, and business groups stripped most mandatory provisions.
The final version requires disclosure of AI use in hiring and insurance decisions but removes requirements for independent audits, third-party testing, and liability provisions that would have made companies responsible for AI-caused harm.
Why it matters: Colorado was supposed to be the template for state-level AI regulation. The watering-down demonstrates how difficult it is to pass meaningful AI regulation when the tech industry has unlimited lobbying resources. The bill that passed is better than nothing, but it’s not the regulatory floor that advocates hoped for. This is the pattern we’ll see across every state: industry pushes for voluntary disclosure over mandatory oversight, and usually wins.
Sources: Denver Post, TechRepublic, Colorado General Assembly
6. 🤖 Poppy: A Proactive AI Assistant That Reorganises Your Digital Life
The story: A new startup called Poppy (from Second Nature Computing) debuted a proactive AI assistant designed to help users organise their digital lives — but with a twist: instead of waiting for commands, it autonomously identifies tasks, organises files, suggests reminders, and even reconfigures app layouts based on observed behaviour.
The assistant runs locally on-device (privacy-focused) and uses what the company calls “ambient intelligence” — learning your patterns without explicit training. It’s designed to solve the “app fatigue” problem where phones have become overwhelming portals of distraction.
Why it matters: We’re seeing the early signs of a shift from “react” to “proact” in AI assistants. ChatGPT, Gemini, and Claude all wait for you to ask. Poppy represents a design philosophy where AI observes and acts without being prompted. This raises obvious privacy questions, but it also points to where assistants are headed: not tools you use, but environments you inhabit.
Sources: TechCrunch