Read this for your industry
Optional. Choose your industry and this piece opens with a situation from your world and closes with a first move for you. Your choice carries across the guides and analysis on this site.
You sit on the board of a licensee
ASIC's Report 798 found that AI governance at licensees was lagging behind adoption, and Hall & Wilcox's summary notes that about 30% of the use cases ASIC reviewed relied on models built by third parties. Directors are approving tools their organisations did not build and cannot fully inspect.
A board can still make sound decisions about these tools by reading the building report. That means asking what each agent does, what quality bar applies and who answers for the result.
What this means for you
- Ask for an inventory of AI use cases and their owners.
- Ask how each tool's output is checked before it reaches a client.
- Ask what happens when a third-party model changes.
You approve technology that goes into every store
Bunnings used facial recognition in 63 stores to protect staff and reduce crime. When the tribunal reviewed the case, it accepted the crime-prevention purpose and still upheld findings that customers had not been properly notified and that governance had fallen short.
The leaders who approved the system meant well. Fluency would have meant asking the questions a building report answers: what it does, what could go wrong and who checks.
What this means for you
- Ask what customer or staff information each AI tool collects, and whether customers are told.
- Ask for the risk assessment before a tool goes into stores.
- Ask who in each region is accountable for how the tool is used.
You sign off tools that may be medical devices
A TGA review found that some digital scribes may meet the definition of a medical device, and the RACGP has added AI to its practice standards, including expectations for governance and consent.
Health leaders can judge these tools well by asking three things: whether the tool is regulated, how clinicians check its output and who answers for it.
What this means for you
- Ask vendors for the regulatory status and intended use of every AI tool.
- Ask how clinicians review output, and how errors are reported.
- Ask what the tool does with patient information.
You fund AI tools for engineering or sourcing teams
The 2025 DORA report describes AI as an amplifier that magnifies the strengths of high-performing organisations and the problems of struggling ones. DORA found that the greatest returns come from the system around the tool, including the platform, the workflows and how teams work together.
Fluent engineering and sourcing leaders ask about that system: the tests, the platform and the quality of supplier data. Those answers tell them whether a tool will help.
What this means for you
- Ask how agent output is tested before release.
- Ask which measures show the tool is helping customers, such as stability, lead time and defects.
- Ask which parts of the platform need work before AI can help.
You choose software you cannot fully check
Installers are adopting AI tools for design, quoting and compliance at the same time as the regulator is finding 60.8% of inspected installations substandard. The owner who buys the tool remains accountable for the work it shapes.
Fluency here means asking how the tool knows the current rules, how its output is checked and what it does when a site does not fit the template.
What this means for you
- Ask vendors how they keep rules such as rebate factors and labelling requirements up to date.
- Ask how the tool flags a design it is unsure about.
- Ask who in your business checks the output, and when.
Reading the building report
Most people who buy a house get a building and pest report first. You do not need to be a builder to read one. You need to know enough to tell a cosmetic crack from a structural one, to ask what the rising damp will cost to fix, and to know when to get the inspector back for a second look.
Without that, you end up trusting whoever sounds most confident: the agent selling the house, or the friend who says every old house is a money pit.
AI fluency for leaders is the ability to read the building report. It is well short of technical expertise, and it is enough to make a sound decision.
The questions fluency lets you ask
With enough working knowledge, a leader can ask:
- What work are we trying to change?
- Where is AI assisting a person, and where is it effectively acting on its own?
- What quality bar applies before anyone relies on the output?
- Where does human judgement need to stay close to the decision?
- Who is accountable when part of the work is machine-produced?
A few years ago these were questions for the technology team. They have become everyday leadership questions, because the answers decide how work gets done.
Why it matters now
AI is moving from something a person uses to something that changes how work is organised. Microsoft’s Work Trend Index describes a shift towards teams of people and agents, with 81% of leaders expecting agents to be moderately or extensively integrated into their company’s AI strategy within 12 to 18 months.
The gap is real at the top. In Deloitte’s second Governance of AI survey of 700 board directors and executives across 56 countries, 66% said their boards still have “limited to no knowledge or experience” with AI. That is an improvement on 79% in the previous edition. It still means most boards are approving AI investment they cannot fully test, a bit like signing a contract on a house without reading the report.
Leaders also shape adoption across the organisation. BCG found the share of frontline employees who feel positive about generative AI rises from 15% to 55% with strong leadership support.
What happens without it
When leaders lack fluency, the most confident voice tends to decide: the vendor demo, the internal enthusiast, the sceptic, the risk committee or the latest productivity headline. Each of those voices has something useful to say. None of them should make the decision alone.
With enough fluency, leaders can make calmer calls about where to experiment, where to slow down, where to redesign the work and where judgement should stay close to the decision.
How to build it
The quickest way I know is to use AI on your own work every week, on something that matters. Draft the board paper, test the forecast, research a competitor. Then check the result properly.
You will learn where it is strong, where it bluffs (see the confidence trap), and which controls you would want before your team relied on it. It is how most of us learnt to read a building report: by reading a few, asking the inspector awkward questions and slowly learning which cracks matter.
At the next board or risk committee meeting, ask management to present three agent use cases against the questions in this piece.
Pick one AI tool used in your stores. Ask the three questions in this piece, and see whether anyone can answer them in writing.
List the AI tools in use across your sites, and ask each vendor to confirm the regulatory status of their tool in writing.
Ask each team lead to name one weakness AI has amplified in their area. Fund that fix before you fund more licences.
Before you renew or buy an AI tool, test it on three of your hardest past jobs and compare its output with what your best designer did.
Questions readers ask
How technical does a leader need to be?
Less than most people fear. You need to know what AI does well and badly, how to tell grounded output from a guess, and what sensible controls look like.
How do leaders build AI fluency quickly?
Use it on your own work every week, on a real task with a real standard, and check the result properly. An hour a week of hands-on use teaches more than a briefing pack.
Sources
- Governance of AI: A critical imperative for today's boards, 2nd edition, Deloitte, 2025
- 2025: The year the Frontier Firm is born (Work Trend Index), Microsoft WorkLab, April 2025
- AI at Work 2025: Momentum Builds, but Gaps Remain, BCG, June 2025
The views here are my own, not those of my employer or its clients.