Article By Sull McIntyre, Chief Product Officer, Wiise
Aussie manufacturers have spent the past three years being told that artificial intelligence will upend everything.
Apparently, it will write reports, predict demand, optimise supply chains, eliminate paperwork and make every employee more productive. Depending on which tech conference you attend, it may also solve world hunger and teach your dog to speak French.
The reality is considerably less exciting and far more useful.
AI is already delivering value for businesses. The problem is that many organisations are discovering that productivity gains and cost control do not automatically arrive simultaneously. Phase 1 of the AI boom was experimentation. A few enthusiastic employees opened accounts, tried some tools and produced some impressive demos. Phase 2 is now underway as companies try to operationalise AI across entire organisations. And that is the bit where things get knotty.
You may have read some of the growing discussion about the cost of large-scale AI usage – what people are in tech calling “tokenomics”. It goes like this: the more value organisations see in AI, the more they use it but the more they use it, the more they spend. What started out as a modest software bill may quickly balloon and become into a material operational cost issue.
This should sound familiar to manufacturers managing soaring energy bills, labour deficits and supply chain volatility. So, how best to avoid the same issues with AI and avoid some of the pitfalls?
The first mistake I see people make is assuming every problem requires the most powerful AI model available. Manufacturing businesses have always understood that different jobs require different tools (no one buys a JCB excavator to dig their backyard flower bed).
AI should be approached the same way. Some tasks genuinely justify advanced models: analysing complex financial information, forecasting demand across multiple variables or identifying emerging supply chain risks may warrant buying the best tech on the market. Other tasks do not. Checking invoice classifications, validating receipts, monitoring routine transactions or handling repetitive administrative processes can often be performed effectively by smaller and significantly cheaper models.
The second mistake tripping people up is weak AI governance. Many firms have little visibility into which AI tools are being used across their business – so-called “shadow AI”. Employees are experimenting with different platforms, uploading documents and connecting data sources with minimal oversight. That may have been acceptable when AI was still a curiosity, but now it is becoming a part of critical business operations. AI has become a major data security risk.
The good news is that manufacturers already understand the importance of process control. They monitor production lines and enforce strict safety procedures and quality controls. It’s in their DNA.
AI requires some of the same discipline. That doesn’t mean manufacturing leaders should crush experimentation; some of the best ideas come from curious employees exploring new technologies. But leaders need to understand which tools are being used and what information is being shared.
The final mistake is to imagine that new technology will solve all your underlying business problems. The manufacturers I see getting meaningful results from AI are combining smart tech with operational expertise, governance and a clear view of what success actually looks like.
Manufacturing has always rewarded practical thinking over fashionable ideas. Artificial intelligence should be treated no differently. The technology is real and the opportunities are real. The invoices will be real too.
Smart manufacturers will prepare for all three.
