The noise around AI right now is all about employment threat – roles disappearing (GitHub, Cloudflare, CitiGroup etc) and the shape of work changing. It remains to be seen how real this is versus business restructure-AI washing – but if you’re running a make-to-order business, from a manufacturing perspective, this entire debate is noise.
In make-on-demand, skills shortages and an ageing workforce demographic (have a read of the Printing United 2025 Industry Census for on this topic) are colliding with growing operational complexity. That collision is where the real pressure sits.
Whether you’re an apparel-deco shop or manufacturing print-on-demand photo-products, the number of SKU production combinations – product × options × print × finishing × packaging × shipping – can not only run into the hundreds of thousands but the number is continuing to grow with innovation and ongoing market fragmentation.
This means the overhead of training and upskilling staff – on machine functions, procedures, maintenance, edge cases, and the special rules that apply for different garments, substrates, customers, or error codes (let alone some of the creative commitments that the sales team make) – is very high.
Yet most businesses don’t have the procedures for it.
Instead most organisations are held up by tribal knowledge. A few people know all the weird rules, all the exceptions, all the workarounds.
When one of them takes annual leave, gets sick, or leaves – or when a seasonal spike hits and you’re onboarding casuals – you’re at risk. And as the Printing United data has shown, that risk is sharpening as the workforce ages. The problem is going to get bigger.
So bringing this back to AI. AI and natural language processing are particularly good at solving this problem and can have a real and immediate impact.
It’s not an exotic application of AI, but it’s doable and can be executed now.
Here are two examples — features we’ve just added to ZenSmart, designed to support people at the order and machine level.
AI-Insights lets you load any company document, training guide, machine manual, or safety plan into a chat interface that’s accessed from where all the normal production processing happens. Ask it anything – production questions, procedures, the lot. The objective is tribal knowledge on demand.
AI-Explain-This lets a team member ask a question about what’s on their screen right then. A VIP order with special conditions. A complex order they’re confused by. A routing instruction that doesn’t track. An error code they don’t know. Instant clarity, right where they need it.
The features themselves are conceptually straightforward. What matters is the job design outcome: people who are less trained can get more capable faster. Onboarding is quicker. Staffing becomes more resilient when someone’s out. Your best people stop being mobile encyclopedias and can focus on what they’re actually good at.
Culturally, better-informed people feel better about their jobs too.
So maybe the conversation shouldn’t be about imagined threats to roles. Maybe it should be about helping the people you’ve got perform better at the roles they’re in right now.
