AI can help solve a maths problem that had sat unsolved for 50 years. It can also get a simple pipe puzzle wrong. Brendan and Ross catch up with Adam Spencer at IT Nation ANZ to explore what that contrast means for businesses, human expertise and the way we adopt new technology.
Watch the conversation
Watch the conversation on YouTube
The short version
AI is most useful in business when it supports a clearly defined task, has access to the right context and remains subject to human review. More capable models can still make basic mistakes, so people need to set the direction, check the result and protect the information involved.
What’s covered
Six things worth taking away
AI becomes more useful when the problem is narrow, the question is specific and the task has a clear goal.
A powerful model can still make basic mistakes. Human review belongs in the process.
AI can give skilled people more options to explore. It does not remove the need to understand the work.
More AI usage does not automatically create more business value. The use case and the result still need to make commercial sense.
As AI produces more polished work, businesses may need better ways to assess practical skills and understanding.
Connected devices create privacy and security questions. Check their settings and keep them separate from sensitive systems where appropriate.
Adam shares a striking contrast from the world of mathematics. He describes a multi-agent AI system helping to tackle a problem that had remained unsolved for decades. He then contrasts this with a much simpler water-and-pipes puzzle that several general-purpose AI platforms answered incorrectly.
The point is not that AI is useless. It is that capability depends on the task, the information provided and the way the question is framed.
AI may misunderstand the problem, miss an important detail or follow an assumption that a person would have recognised immediately.
A broad request such as “How can we improve our business?” is unlikely to produce a useful answer on its own. A more focused request gives AI something practical to work with:
AI can be impressive and unreliable at the same time. The task still needs a person who understands the work.
Adam describes a multi-agent system using a detailed, two-page prompt to work through a difficult mathematics problem.
The system was divided into around 64 agents. Some were trying to solve the problem, while others were challenging the proposed approaches. It was also told to keep exploring instead of stopping at the first possible answer.
Most businesses do not need dozens of AI agents working on advanced mathematics. They do need the same discipline when applying AI to everyday work.
Before introducing a new tool, ask:
The conversation also explores what AI means for human expertise.
Adam describes AI as a way for skilled people to explore more ideas, test possible directions and spend less time pursuing options that are unlikely to work.
A mathematician still needs to understand the problem. An operations leader still needs to know how the business works. A finance professional still needs to judge whether an answer makes commercial sense.
AI can help people investigate further and move faster. It cannot take responsibility for the decision.
Adam uses the phrase “brain fade” to describe the risk of becoming too dependent on general-purpose AI tools.
The concern is that people may produce polished work without remaining connected to how it was created or whether it is accurate.
Businesses should keep people involved in the work AI supports. That may mean reviewing sources, explaining recommendations, checking calculations or demonstrating practical skills.
The conversation ends with a surprising story about a robot vacuum.
The device was connected to the internet, had cameras and stored images remotely. When someone found a way to access the system, they were able to see far more about connected homes than the owners may have expected.
The same principle applies to business AI tools and connected devices. Before introducing them, organisations should consider:
Before introducing AI
Choose one process that the business wants to improve. Document how it currently works, set boundaries around the information involved and decide what a useful result looks like.
Then ask:
A practical model for AI adoption
About the guest
AI keynote speaker and commentator
Adam speaks with organisations across different industries about AI and what it means for people, work and business. In this conversation, he brings that broad view to the rapid changes he has seen over the past 12 months.
Common questions
Want to talk about where AI could fit in your business?
IT Consulting
IT Strategy
Data Services
IT Infrastructure Projects
IT Procurement
Virtual CIO
Offshore Staff Outsourcing (BPO)
Managed AI & IT Services
IT Support Services
Managed Cloud Services
Managed SharePoint Services
End User Computing
Unified Communications
Connectivity
Managed Security Services
Cybersecurity Risk Assessment
Business Continuity Planning
Cloud Backup and Disaster Recovery
Automation
Digital Transformation
Power BI
Careers
Awards and Certifications
Management Team
Case Studies
Sell Your IT Business
Our Commitment to the Environment
CSAT Charity Donations
Media Kit
Auckland | Tāmaki Makaurau
