Blog
AI automation insights
Practical guides for Australian businesses navigating AI automation. No jargon, no hype — just what works.
September 2026 • 7 min read
Claude and OpenAI's Misalignment Reports: What to Ask
OpenAI now publishes reports of its models misbehaving. What Australian Claude users should take from it, and the safety questions to ask any AI vendor.
September 2026 • 7 min read
Salesforce in Claude: What Changes for Sales Teams
Salesforce in Claude brings pipeline, accounts and 37 sales skills into Claude with seller-approved CRM updates. What it does and what to check first.
September 2026
AI-Native Revenue Teams: Rolling Out Claude Across Sales
How to roll out Claude across a sales team, based on Anthropic's revenue organisation guide: Cox's 7x ROI, Cyera's 88% adoption, and an Australian plan.
September 2026 • 6 min read
Claude vs Gemini Voice AI: Reading Gemini 3.8 Live
Gemini 3.8 Live leads on voice quality, but completed 35.1% of banking tasks. What Australian businesses choosing a voice agent should weigh first.
September 2026 • 6 min read
Claude Users and Google's 86% Everyday-AI-Use Stat
Google found 86% of conversational AI use happens outside work. What that means for Australian businesses whose staff bring home AI habits to work.
September 2026 • 6 min read
AI Pilot to Production: Claude's Four Key Decisions
Only 23% of leaders report sustained AI impact. The four decisions from Claude and Accenture's pilot-to-production guide, applied to Australian firms.
September 2026 • 7 min read
Claude for Financial Advisors: An AU Practice View
Claude's new plugin for financial advisors bundles US connectors and skills. The control model is what Australian advice practices should copy.
September 2026 • 6 min read
Claude Code CI Scaling: What 25x Job Growth Broke
When Claude wrote 80% of Anthropic's code, CI jobs rose 25x in six months. What broke, the three patches that failed, and what AU teams should plan for.
September 2026 • 6 min read
Open Source AI in Childcare: Start With the Child Data
For Australian childcare and early learning providers, the AI question isn't open or commercial. It's whether child data touches the model at all.
September 2026 • 5 min read
Small Language Models Cost Less: 8B vs 70B for AU SMBs
Small 8B open-weight models now handle narrow, repetitive tasks cheaply. Where Australian SMBs save by routing them, and where Claude should stay.
September 2026 • 7 min read
Claude Tag in Healthcare, Without Patient Data
Three healthcare companies run Claude Tag in Slack without it seeing patient data. The channel and connector scoping behind it, for AU practices.
September 2026 • 6 min read
Claude for Small Business Australia: 43 New Workflows
Claude for Small Business added 43 workflows and 27 integrations, including Xero. What the September 2026 update means for Australian owners.
September 2026 • 6 min read
Open Source AI for Franchise Networks: Who Owns the Data?
In most Australian franchise networks, head office doesn't own the customer data. What that means for a central AI model, and what is safe to centralise.
September 2026 • 5 min read
Sakana AI Fugu: A New Open-Weight Lab for Your Vendor List
Sakana AI shipped Fugu Ultra v2.0 and Fugu Max on 11 September 2026. Why a new open-weight lab changes your vendor register, not your stack.
September 2026 • 6 min read
3 Claude Fable 5.1 System Prompt Rules Worth Stealing
The viral line from the Claude Fable 5.1 prompting guide was about prose. The three rules that cut agent rework are about stopping, tools and memory.
September 2026 • 6 min read
Evaluating Claude Code: Lessons From an AI-Resistant Test
Anthropic rewrote its hiring test three times as Claude kept passing it. What that teaches about evaluating Claude Code on your own team's work.
September 2026 • 6 min read
Claude Fable 5.1 vs DeepSeek V4 Flash: Real Cost
DeepSeek V4 Flash wins on list price. Claude Fable 5.1's cheaper cache narrows the gap on repeat-context work. How to run the real AUD numbers.
September 2026 • 6 min read
Ministral-3-3B-Instruct-2512 for Sites With No Signal
Mistral's 3B multimodal Ministral model runs on a tablet with no internet. Where it fits for Australian site crews, what it can't do, and what it costs.
September 2026 • 6 min read
The HBM Shortage and Open-Source AI Cost: Plan Ahead
Cheap open-weight tokens rest on a temporary glut of high-bandwidth memory. How to stress-test an Australian GPU budget before that supply tightens.
September 2026 • 6 min read
Open Source AI for Australian Recruitment and HR Agencies
Self-hosting an open-weight model keeps CVs off third-party APIs, but it is not the Privacy Act answer on its own. Tasks, risks and AUD costs.
September 2026 • 5 min read
Tencent Hy4: A 770B Apache 2.0 Model and Your Vendor List
Tencent's Hy4 replaced Hy3 in under two months. What a 770B Apache 2.0 open-weight model means for Australian vendor lists and licence reviews.
September 2026 • 6 min read
Essential Eight and Open-Weight AI: The Missing Map
Self-hosting an open-weight model is not only a data residency question. How the Essential Eight maps onto your inference server, and what it costs.
September 2026 • 6 min read
Token Minimisation for Open-Weight LLM Deployment
In 2026 the teams running open-weight models at volume stopped sending more tokens and started sending fewer. Why that skews most Claude cost comparisons.
September 2026 • 6 min read
Open Source AI for Australian Vet Practices: Privacy First
Open-weight models can draft after-hours triage replies and tidy consult notes for vet clinics. Where they help, and where Privacy Act questions bite.
Get AI automation insights in your inbox
One short email per month. What's working in Australian AI automation. Unsubscribe anytime.