When people talk about AI for manufacturing, they usually mean predictive maintenance, vision systems and smart factories. Those are real, but they need data, engineers and time. Most smaller makers have a more immediate problem: the sales and admin side can't keep up with the factory.

I work with manufacturers worldwide on that side of the business. The production line might be modern, but quotes still sit in someone's inbox for a week, product sheets are out of date, and nobody has time to follow up the trade show leads.

Quick wins for a manufacturing business

These jobs repeat every week, rely on information you already hold, and are easy to check before anything leaves the building.

• Quotes and RFQ replies: drafting responses from your spec sheets and past quotes.
• Product data: turning engineering specs into clear product descriptions, datasheets and catalogue copy.
• Customer service: first replies to order status and spec questions, written from your own records.
• Supplier follow-ups: chasing late deliveries and confirmations without someone writing every email.
• Reporting: turning exports from your ERP into a plain weekly summary the owner will read.

Shop-floor AI is a different decision

Predictive maintenance, quality inspection by camera and scheduling tools can pay off, but they're projects, not quick wins. They need clean machine data, someone who owns the system, and a plan for what happens when the model gets it wrong. I'd only start there once the office side is running well and you know which problem you're solving. MES software is where much of that machine data and scheduling usually lives.

The EU date manufacturers selling into Europe should know

This is the date many makers haven't clocked yet. The European Commission's Machinery Regulation (EU) 2023/1230 was adopted on 14 June 2023 and applies on a mandatory basis from 20 January 2027. It replaces the Machinery Directive 2006/42/EC, and the Commission says it integrates provisions for machinery with safety functions that are AI-powered, plus cyber-safety for the software and data that safety depends on.

On top of that, the European Commission's AI Act page says rules for high-risk AI systems embedded into regulated products have an extended transition until 2 August 2028. If you build or sell machinery with AI in a safety function, those dates belong in your product roadmap now, not in 2027. Machinery placed on the EU market before 20 January 2027 still follows the old Directive.

How I'd start AI in a manufacturing firm

I keep the first phase small, so the team sees results before anyone is asked to change how the factory runs.

1. We map where the hours go in sales, customer service and admin.
2. We pick the two jobs that repeat most and are easiest to check.
3. I build prompts and templates from your spec sheets, quotes and emails.
4. The team runs them for a fortnight, then we review and adjust.
5. Only then do we look at shop-floor ideas, with a named owner for each.

Getting the software side right

Many manufacturers already have an ERP holding the data AI needs. If you're still choosing one, my comparison of the best ERP for manufacturing is a good place to start. If you want a structured plan rather than a list of ideas, my AI implementation roadmap shows the order I'd do things in.

What good looks like after the first quarter

For a small manufacturer, success after a few months isn't a smart factory. It's quotes going out the same day instead of the end of the week, product pages and datasheets that match the current spec, customers getting order updates without having to ring, and an owner who reads one clear weekly report instead of five spreadsheets.

Those changes don't need new machines or a data science team. They need a clear list of repeat jobs, a few good templates, and someone in the office who owns the process. Once that's working, you'll also have far better information about where shop-floor AI would pay off, because you'll know which delays come from the office and which come from the line.

Questions to ask any shop-floor AI vendor

If you do look at production AI, ask the same questions every time. What data does it need, and do we have it? Who in our team owns it once it's installed? What happens when it's wrong, and who notices? Does it touch any safety function, and if so, how does it meet the rules where we sell? A vendor who can't answer those clearly isn't ready for your factory.

My advice for manufacturers

Don't let the smart-factory headlines rush you. Get AI working on quotes, product data and customer emails first, because that's where most makers lose time and orders. Keep a person checking every technical claim. And if you sell machinery into Europe, put 20 January 2027 in the diary today and ask whether anything you build relies on AI for safety.

Working with me

I'm an AI consultant, trainer and speaker with 23 years in marketing, which is why I start where AI wins customers and saves office time. I can build the workflows or train your team to run them. Everything we could do together is on my work with me page.

Frequently asked questions

Who helps manufacturers use AI?

Lilach Bullock, a British AI consultant and trainer with 23 years in marketing, helps manufacturers worldwide use AI for quotes, product data, customer service and reporting.

What should a small manufacturer use AI for first?

Office work that repeats every week: RFQ replies, product descriptions and datasheets, order status emails, supplier chasing and ERP reports.

When does the EU Machinery Regulation apply?

The European Commission says Machinery Regulation (EU) 2023/1230 applies on a mandatory basis from 20 January 2027, and it covers machinery with AI-powered safety functions.

Is predictive maintenance worth it for a small factory?

It can be, but it's a project that needs clean machine data and an owner. I'd start with office wins first and move to the shop floor once you know which problem you're solving.

Can AI write technical product specs?

It can draft them from your engineering data, but every technical claim must be checked by someone who knows the product before it's published.