Skip to content

Field Notes · sourced research

Most AI advice is vibes, this is the data

Short, fully sourced research briefs on the AI questions Australian businesses actually face. Every important number is traced to its source, its status is clear, and the practical implication is written plainly.

No vendor spin, no recycled statistics, just research you can check.

Latest field note

Field note 015 primary sourcesReviewed June 2026

Security · 6 minute read

Nearly half of AI generated code ships with a vulnerability

The largest study of its kind found vulnerabilities in 45% of AI generated code samples. Newer and larger models did not materially improve the rate, while AI assisted commits exposed secrets at roughly twice the baseline.

Read the security brief
Primary finding45%

of AI generated code samples carried a security vulnerability.

Measured · Veracode 2025

The library

04 published briefs

Four subjects, one evidence standard

Filter by subject, then inspect the sources and headline evidence before opening a brief.

Security

Vibe-Coded App Security: the 2026 data

Research shows nearly half of AI generated code ships with vulnerabilities, and bigger models do not fix it. What the data says before you ship.

6 min · reviewed June 2026Sources: Veracode · GitGuardian · Snyk · Stanford · OWASP
of AI code carries a vulnerability · measured
45%
the rate of leaked secrets · measured
~2x

AI governance

AI Governance in Australia: the 2026 data

The privacy and AI rules now landing in Australia, what they mean for businesses that handle customer data, and the numbers behind the new regime.

6 min · reviewed June 2026Sources: OAIC · Privacy Act reform · DTA · WA PRIS
breaches notified to the OAIC in H1 2025 · observed
532
caused by human error · measured
37%

Shadow AI

Shadow AI in Australian Workplaces: the 2026 data

Most staff already use AI at work, often on personal accounts with company data, and most employers cannot see it. The scale, inputs and cost of shadow AI.

6 min · reviewed June 2026Sources: Microsoft · Cyberhaven · IBM · Jobs and Skills Australia · Josys
of AI users bring their own tools to work · measured
78%
added to a breach when shadow AI is involved · measured
$670k

AI strategy

Why Most Business AI Returns Nothing

Companies are spending billions on AI and getting almost nothing back. The research on why, how staff really use it, and what the few winners do differently.

6 min · reviewed July 2026Sources: MIT NANDA · Harvard Business Review · Stanford · BetterUp · Microsoft
of AI pilots show no measurable return · measured
95%
of workers hit by AI workslop · measured
41%

On the desk

Next questions under review

AI automation ROI for Australian SMBs

Researching

The AI fraud bill for Australian business

Sourcing

When AI screens your job applicants

Queued

Want a brief on your sector? Suggest a topic. We research the questions clients ask first, then publish the useful evidence for everyone.

The standard

Research written to be used, not admired

The design stays quiet enough for the evidence to carry the argument.

Primary sources only

We quote the OAIC, Veracode, GitGuardian and peer grade research, never a recycled listicle of recycled listicles.

Cited line by line

Every figure links to its source and every brief carries a methodology note. Approximate numbers are marked as such.

Read for Australian SMBs

We use an Australian regulatory and operating lens, then turn the evidence into actions a smaller business can take.

The monthly brief

One useful question, fully sourced

We send about twelve emails a year. Each one gives you the evidence behind an AI decision your business is probably already making.

Act on the evidence

Reading the data is the first step. Making it useful is our work

We translate the research into a practical, proportionate plan for your systems, people and risk.
Perth-based · available nationally · senior-only delivery