AI Slop vs Useful AI: A Plain English Guide for Industrial Businesses

If you've scrolled through LinkedIn or Facebook lately you'll have seen it: the glossy poster where the engineer in the hard hat has six fingers, the post that opens with "In today's fast-paced world" and then goes nowhere for 400 words, and the video advert where a suspiciously perfect actor tells you, with far too much enthusiasm, that a piece of software changed their life. It's become known as AI slop, and if you run an engineering or manufacturing business there's a fair chance it's left you thinking the whole thing is a gimmick you're better off ignoring.

We understand that reaction, because we've spent a good chunk of the last couple of years wincing at the stuff ourselves (and producing a bit of it while we were learning). We've also used the same tools to build software we could never have afforded to commission, though, and that experience has changed how we think about AI in a smaller business.

This post covers both sides of the coin: what slop actually is and why there's so much of it; what's changed underneath all the noise; what we built ourselves; and a practical guide to using AI in your own business without adding to the pile.

What we mean by AI slop

Slop is the name that's stuck for content made with AI that nobody really checked, thought about, or cared about, and in our experience, it shows up in three main forms.

Visual slop

Visual slop is the stuff you spot in half a second, with the same shiny, over-lit style on every poster, hands with the wrong number of fingers, text in the background that's nearly but not quite English, and AI actors in video adverts whose faces don't move quite the way faces should. Facebook is full of it, and so (sadly) are plenty of company websites that swapped their real photos of real kit for something generated in a few seconds.

Written slop

Written slop is harder to spot, but you'll spot it after a few paragraphs, because it's long, confident, and says almost nothing. It loves words like "landscape", "pivotal", and "seamless"; it groups everything into neat threes; and it ends with a sentence about the future looking bright. We've written before about why plain English matters on your website, and slop is about as far from plain English as you can get, since most of it is padding dressed up as expertise.

Code slop

Most business owners never see code slop because it hides behind a website or app until something breaks. It's code that looks like it works and passes a quick test, but underneath it's insecure or held together by assumptions nobody checked, and it tends to fall apart the first time a real user does something unexpected.

Why there's so much of it

The simple explanation is that AI makes it quick and cheap to produce something that looks finished, and many people stop there. You type in a one-line request, get back a polished-looking result in ten seconds, and because it looks the part, you publish it without asking whether it's accurate, whether it sounds like you, or whether anyone actually needs it.

We'd argue the tools themselves aren't really the problem. If you gave a capable copywriter a one-line brief, no background, and no editing, you'd get something fairly generic back, and AI works the same way, only faster and on a far bigger scale. Slop is what you get when no real knowledge goes in, and nobody with the right expertise checks what comes out.

There's also a volume problem. When a blog post or a poster costs next to nothing to produce, some businesses decide the answer is to produce hundreds, and that's how you end up with a feed full of content that all looks and sounds the same. Google has said fairly clearly that it cares about whether content is helpful and original, not how it was made, which tells you where the bar sits, and our guide to writing SEO-friendly content covers what that means in practice.

The bit that gets lost in the noise

So, here's the thing: underneath all the slop, something useful has happened: the cost of making things has dropped through the floor. Design work, written work and software that used to need a specialist and a few weeks of waiting can now be produced (at least to a decent first draft) by someone who understands the problem, even if they can't write code or lay out a brochure themselves.

For a large business with an in-house marketing and IT team, that's a handy efficiency gain, but for a 30-person engineering firm in Aberdeen or Stoke it's a much bigger deal. Things that were previously out of reach, like a bespoke internal tool or a properly written set of product pages, are now realistic without hiring anyone new or waiting for a gap in an agency's schedule.

That's what we mean by saying AI has democratised production. Expertise and judgement are still in short supply, which is why there's so much slop, but if you've got the expertise (and most engineering businesses have bags of it), AI makes it much easier to turn that knowledge into finished work.

What we built with it ourselves

The best example is our own. Like most agencies, we need quick, reliable answers to questions like "how much of this client's retainer have we used this month?" and "have we got enough people to get through next month's work?", and getting those answers used to mean pulling information together from several places by hand.

We have developers on the team, but a few years ago a bespoke system to fix that would still have been hard to justify, because the development time would have come straight out of client work. AI changed that: one of our most experienced developers could use it to handle much of the routine coding and build the system far faster, and now the whole agency runs on it. We call it REDMan.

REDMan logs every piece of work we do for each client against their retainer, keeps a running balance of whether we've over or under-delivered, and produces our monthly client reports. It also reads holidays from the team's calendars to produce a monthly workload-versus-capacity report, so we can see well in advance whether we've got more work than people, and it pushes tasks straight onto our project boards in Monday so nobody has to type them in twice.

The important part is who was in charge. AI sped up development, but a developer with years of experience did the work, knew how the system should be put together, could spot when the AI produced code that looked right but wasn't, and tested everything against our own figures before we relied on it. The rest of us brought the twenty-odd years we've spent learning how an agency actually runs, and we were very specific from the start about what REDMan needed to do.

If we'd handed the same job to AI without an experienced developer running it, we'd most likely have ended up with code slop, a tool that looked impressive and gave the wrong answers.

How to use AI without adding to the slop pile

If you're an MD wondering whether any of this applies to your business, we suggest this approach, based on what's worked for us and what we've seen go wrong for others.

1. Start with a boring problem

Start with something dull and repetitive that already eats up time, like reconciling a spreadsheet nobody trusts or writing the same report every month. Internal jobs are forgiving: if the first attempt is rough, only your team sees it, and the time you save is easy to measure.

2. Give it your knowledge, not a one-line brief

AI knows a little about almost everything and nothing at all about your business, so the quality of what comes out depends heavily on what you put in. Give it your product specs, past proposals, the terminology your customers use, examples of work you're proud of and examples of work you'd never send, and you'll get something much closer to usable on the first pass.

3. Keep an expert in the loop

Whoever checks the output needs to know the subject well enough to spot when it's wrong, because AI can be wrong while sounding completely sure of itself. For technical content, that means an engineer reads it before it goes anywhere; for anything involving numbers, it means someone checks the sums against the original source.

4. Never publish the first draft

Treat whatever AI gives you as a first draft from a keen junior who's read a lot but has never set foot on site, so it needs fact-checking and rewriting in your own voice before anyone outside the business sees it.

5. Be careful what you share

Check what a tool does with your data before you paste in drawings or pricing, because some free consumer versions may use what you type to improve future models. The business plans from the main providers generally offer better protection, but it's worth reading the terms (and checking your client contracts and NDAs) before anything sensitive goes in.

6. Keep the human stuff human

Real photos of your team and your workshop will always beat a generated image, because buyers want to see who they're actually dealing with. The same goes for case studies and testimonials, which need to come from real projects and real people, and if you're thinking about the bigger picture, our post on content marketing for engineering companies explains why that kind of first-hand material does so much of the work.

Where engineering and manufacturing firms could start

Every business is different, but these are the kinds of jobs where we'd expect AI to save an industrial firm real time without much risk, provided someone who knows the work checks what comes out:

• Turning engineers' rough notes into first-draft tender responses or method statements
• Rewriting technical product data into clear web pages for buyers who aren't specialists
• Building small internal tools, like a quote calculator or a job tracker, that you'd never have paid a developer for
• Summarising long specifications, standards or contracts so you can get to the clauses that matter
• Digging through data you already have, such as enquiry sources or sales figures, to spot patterns

We'd hold back on anything customer-facing that goes out unchecked, generated imagery pretending to be your real facility, and bulk content produced purely to fill a website, because that's where trust gets damaged, and you start to look like everyone else. AI is also changing how buyers find suppliers in the first place, which we've covered in AI search and what engineering companies need to know.

So, is AI a force for good?

In our opinion, yes, with a fairly big caveat attached. The slop is real and annoying, and plenty of businesses are quietly damaging their reputations with it, but writing AI off because of the worst examples is like writing off email because of all the spam.

Jeff Goldblum's character in Jurassic Park complains that the scientists "were so preoccupied with whether or not they could, they didn't stop to think if they should", and that's a fair summary of most slop. The businesses getting real value from AI are the ones that put their own expertise in and check everything that comes out, and for a smaller industrial business that can mean doing things that were simply out of reach a few years ago.

If you'd like to talk through where AI might (or might not) fit in your business, we're happy to have a 15-minute chat with no hard sell, and if we think you're better off leaving it alone for now, we'll tell you. You might also find our thoughts on the value of using a digital marketing agency useful if you're weighing up what to do in-house and what to hand over.