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 decide which AI tools your advice business buys
With 15.9 million Australians holding unmet advice needs, the outcome advice businesses want is capacity: more clients served well by each adviser. Tools that remove a specific task, such as file notes or meeting preparation, speak directly to that outcome.
If you are buying or building AI for an advice business, start with the hour each adviser gets back and the client who is seen sooner.
What this means for you
- Buy against a named task and a measurable saving.
- Watch which tools advisers already pay for themselves, because that shows real demand.
- Ask vendors for evidence from practices like yours.
You buy technology for a network of stores
Recorded retail crime in Victoria rose 20% to 99,114 offences, the highest on record. Retailers buy technology against outcomes like that: fewer incidents, safer staff and less loss.
AI products for retail sell when they name the outcome and the store-level job they change.
What this means for you
- Buy against a measurable store outcome, such as incidents, loss or hours saved.
- Ask store managers which tools they would pay for if they could.
- Complete the privacy and safety assessment before the pilot starts.
You decide which AI tools reach your clinicians
AI scribes spread through general practice faster than most clinical tools, with use doubling in a year. They solve one specific job, documenting a consultation, and that focus is a large part of why clinicians took them up.
Health networks buying AI can learn from that. Clinicians adopt tools that give them time back on a named task.
What this means for you
- Buy against a named clinical or administrative task and the time it returns.
- Treat clinicians' own choice of tools as a demand signal, and bring those tools under governance.
- Check regulatory status and privacy before you scale.
You buy AI for engineering or sourcing teams
By 2025, 90% of technology professionals were using AI at work. Much of that adoption started with individual engineers choosing tools for specific jobs, such as writing tests or explaining unfamiliar code.
Buyers of AI in engineering and sourcing can follow that demand. The tools people already use point to the jobs worth paying for.
What this means for you
- Survey teams on the tools they use and the jobs they use them for.
- Buy against a measurable outcome, such as lead time or supplier defects.
- Bring popular tools through proper security review, and keep the ones that pass.
You decide what software your installation business pays for
With more than 457,000 batteries installed in the program's first year and a regulator focused on installation quality, installers need outcomes they can count: faster quotes, fewer call-backs and jobs that pass inspection.
AI products that sell in the trades name one of those outcomes and fit the way crews already work.
What this means for you
- Buy against a measurable outcome, such as first-time pass rate or quote turnaround.
- Ask your crews which tools they already use on their phones.
- Trial a tool on your own jobs before you commit.
The Weet-Bix test
A box of Weet-Bix tells you what is inside, what it costs and what you will get for breakfast. Nobody in the supermarket aisle needs to understand the milling process to decide whether to buy it.
Much of enterprise AI is still sold the other way round. The pitch starts with the model, the architecture and the benchmark scores, and leaves the buyer to work out what it will do for their team on a Monday morning.
Buyers pay for what the model does: the solved workflow, the reduced pain, the faster decision, the saved effort. The model is one ingredient.
Job-specific products are easier to buy
A general “AI skill pack” asks the buyer to work out where it fits. A tool that cuts a supply chain lead’s weekly exception review from a day to an hour answers that question on the front of the box.
The research on buying points the same way. Fortune’s write-up of MIT NANDA’s State of AI in Business 2025 reported that buying AI tools from specialised vendors and building partnerships succeeded about 67% of the time, while internal builds succeeded about a third as often. Specialist vendors have usually already done the hard part of fitting AI to a particular job.
Shadow AI is a demand signal
The same research found what Fortune called a “shadow AI economy”. Workers at more than 90% of companies use personal chatbot accounts for daily tasks, often without IT’s approval, while only 40% of companies have official subscriptions.
It is tempting to treat that, and the scrappy tools people build for themselves, as a risk to shut down, and IT teams have good reasons to worry about data and security. Manage the risk, and study the behaviour as well. People rarely build workarounds unless the pain is real and repeated, so each workaround is a small piece of customer research. Put them together and you have a map of problems waiting for a proper product or service.
For builders, advisers and buyers
Product builders: start from a role, a workflow and a pain point. Keep the complexity out of sight and price on the outcome where you can.
Advisers: short workflow reviews, small pilots and a look at the tools staff already use are good ways to find where the value is. Follow the pain.
Buyers: ask every vendor to describe the outcome in your team’s terms, with the measure you would use to check it. If the answer starts with the model, ask again.
Where the value ends up
The winners in enterprise AI may not have the most impressive model story. I expect them to be the companies that explain their value most clearly, keep the complexity out of the buyer’s way and learn fastest from what customers are already trying to fix for themselves.
AI is becoming a capability inside other products, which makes it no less important and moves the money towards whoever packages the outcome best. The products that win will pass the Weet-Bix test: you can see what you get before you open the box.
Ask your advisers which AI tools they use privately. The most common answer is your next business case.
Write a one-page outcome brief for your next AI purchase, naming the store-level measure and its current baseline, before you talk to vendors.
Ask clinicians which AI tools they already use. Pick the most common one, assess it properly, and decide whether to adopt it across your sites.
Run a short survey of the AI tools your engineers and buyers use. Pick the two most common jobs and write the business case around those outcomes.
Pick the outcome that costs you most today, which for many installers is call-backs. Ask every vendor to show how their tool would reduce it, using your last month of jobs.
Questions readers ask
Does the model choice matter at all?
It matters to the builder. For most buyers it is like the engine in a delivery van: they care that it is reliable and cheap to run, and few ask about the brand.
Should we shut down shadow AI?
Manage the risk, and study it before you act. When people work around official tools, they are showing you where the pain is.
What is a good starting point for an AI product?
One role, one workflow and one pain you can measure. Make the value obvious in that job's terms before you broaden it.
Sources
- MIT report: 95% of generative AI pilots at companies are failing, Fortune, 18 August 2025
- The 'shadow AI economy' is booming: Workers at 90% of companies say they use chatbots, but most of them are hiding it from IT, Fortune, 19 August 2025
The views here are my own, not those of my employer or its clients.