AI is starting to bore me/annoy me.
Or at least the hype around it.
Nvidia and Oracle are printing money. Microsoft is cutting billion-dollar deals.
But inside most companies?
AI pilots are crashing and burning. MIT says 95% of enterprise generative AI pilots flop. WorkOS found 42% of firms scrapped their projects this year.
My take?
It’s not the tech. It’s the data.
AI doesn’t fail because it’s dumb. It fails because your data is.
You can buy the shiniest model in the world, but if you feed it garbage, it spits out garbage.
The most significant problems aren’t “AI limitations” — they’re boring, human-made messes.
I know because I’ve caused a lot of those messes.
- Messy content set as “knowledge”. If it’s inaccurate or can’t be parsed properly it’ll hallucinate more than a stoner at a party you don’t want to be at
- No curation. zero effort spent organising knowledge into a reliable “source of truth.” I’ve made the mistake of thinking that the “AI” will figure out the mess I’ve fed it. Never the case.
- Ridiculous expectations. Mucking around with AI Assistants and creating custom GPTs. I have no right to be grumpy that the output is shit if my input is.
Informatica’s 2025 survey nailed it.
43% of AI failures come down to data quality and readiness.
Why data is the make-or-break
Think of AI like an apprentice. It learns from whatever it’s handed.
- If your business manuals are outdated PDFs… the AI learns obsolete rubbish.
- If your CRM is full of duplicates and spelling mistakes… the AI thinks you’ve got three different “Acme Pty Ltds.”
- If your website content is generic filler… the AI can’t distinguish your brand from competitors.
AI isn’t magic. It’s data + workflows, in disguise.
The hidden cost of ignoring data
When companies skip the hard slog of data plumbing (good god it’s laborious!), theyyyy…
- Burn budgets on pilots that never scale.
- Confuse staff with tools that don’t match real processes.
- Have one very bad output outweigh the value of ten good ones
- Risk of privacy breaches as unsorted files get surfaced by AI.
- Feed the hype cycle — inflating expectations, then tumbling into disillusionment.
It’s why the hype charts (Gartner’s Hype Cycle, Kübler-Ross’ change curve) feel so familiar: from “shock and awe” to “oh crap, this doesn’t work.”
What small businesses can learn (without wasting the big bucks)
Here’s the part that matters for you. The data you already own is your competitive edge.
For SMBs, fixing that means this…
- Content curation: organise, label, and centralise your docs, posts, and FAQs.
- Start small: automate the back-office grunt work where data is easiest to structure (invoices, forms, customer queries).
- WordPress as your data hub: instead of letting AI chew on random silos, use your website as the clean, searchable foundation.
- Buy smart, don’t build: pick tools that work with what you’ve got; don’t reinvent the wheel.
At Loud Cow, this is where we start with clients.
Clean up the content first. Then add AI.
Closing thoughts…
Poor old data isn’t sexy, but it’s the difference between an AI project that saves time and one that becomes another forgotten pilot in the corporate graveyard.
So before you splash out on shiny models or “AI-powered” subscriptions (omg – like meeee!), ask – is our content organised, current, and findable?
Because AI isn’t failing 95% of the time.
Your data is.
