19/10/2026
This week I crossed to California to chat with Jeff J Hunter from AI Money Group.
We discussed to AI news stories:
1. New research showing strong return on AI investment is more than three times as likely when teams redesign their workflows rather than adding AI to existing tasks.
And
2, Anthropic warning that AI could change future computer viruses to adapt while they spread
On getting value from AI:
Jeff’s view is that time saved was never a great measurement.
When AI first arrived, that is what everyone reached for.
His focus is revenue-generating activity.
Automating more marketing through social media, through emails.
Once AI at least breaks even against what you spend, adoption gets much easier.
My own research shows most people try AI to save time. Jeff said he did too.
What actually happened was realising how many more things you can now do. You end up chatting back and forth with AI about everything you have always wanted to do.
His measure: faster, better and cheaper. The trifecta.
He argues we have never had a technology that delivers all three at once.
On easing into AI agents:
Jeff’s analogy is like learning to drive. The student is behind the wheel.
The teacher sits in the passenger seat with their own brake pedal.
As you give AI more access, keep your foot over that brake.
What NOT to do first:
Integrating it into your social media, your emails and all your systems at once is the worst possible start.
People can send emails that get your agent to do things.
There is a documented case of a security consultant at Meta whose agent deleted 3,500 emails from their corporate account.
Treat it like a human.
You would never hire someone off the street and hand them the keys to every database you own.
On AI-adaptive malware:
Traditional malware is pre-programmed. If your computer does not have the right conditions, it cannot do much.
With AI, the malware analyses what is happening on your machine first. It finds vulnerabilities. It can replicate itself across your network.
I asked how close this is. Jeff said it is already here.
He expects an attack carried out by AI models before the end of the year.
He also pointed to open source models running without supervision.
What businesses can do:
Be cloud-based.
Have backups of your backups.
Be very controlled about the access you give AI across your network and systems.
Jeff also sees real opportunity in security.
People who were previously hackers coming in as anti-hackers.
Cat and mouse, and whoever has the better development team wins.
If there is one thing to take from this episode, it is access.
Be deliberate about what you hand over.
Thank you Jeff for joining me from California.
University of the Sunshine Coast Dharana Digital
