A business owner sits across from you. Smart, successful, genuinely open to what AI could do for their company. And then they say it:
Look, I've had three of you lot in here already. Everyone's got a deck. Everyone's got case studies. Most of it's bullshit. How am I supposed to know who's actually telling the truth?
We hear this constantly. And honestly? They're right to be suspicious.
The market is broken
Here's the reality nobody in the AI industry wants to say out loud: 95% of generative AI implementations at companies are failing, according to MIT research. Not struggling. Not underperforming. Failing. More than 80% of AI projects fail outright — double the failure rate of traditional IT projects, according to RAND Corporation.
Meanwhile the number of people selling AI solutions has never been higher. The barrier to entry is essentially zero — watch some YouTube videos, build something in a no-code tool, call yourself an AI consultancy, start sending automated LinkedIn messages to anyone with a budget. Thousands of people are doing exactly this right now.
The problem isn't that they exist. The problem is that on the surface, they're indistinguishable from people who actually know what they're doing. And over half of new English-language articles published online are now primarily AI-written, up from roughly 10% before ChatGPT launched in late 2022. The slop isn't just in the AI solutions being sold — it's in all the content being used to sell them.
The spaceship problem
A colleague described it well after a particularly frustrating pitch meeting.
It felt like we were telling someone we had a spaceship behind the trees. And they're looking at us like we're mad — because the last five people who said that turned out to have a car with tinfoil on the roof.
That's the market right now. Real, business-changing AI capability exists. But it's buried under so much noise that the people who need it most have become, reasonably, completely cynical.
We've seen it play out. A company comes highly recommended, we set up a call, the next day they go with their mate instead — cheaper, said he could do it, why not. Three months later the mate's delivered something that barely works. Now they're back.
This isn't unusual. Consulting firm West Monroe describes widespread "AI fatigue" among clients — businesses exhausted by proof-of-concept projects that promised transformation and delivered nothing, often because the wrong tool was applied to the wrong problem.
The cycle keeps repeating. And it will keep repeating until buyers get better at asking the right questions.
Five questions worth asking before you sign anything
We're not going to tell you to check their LinkedIn followers. Followers mean you're entertaining. They don't mean you can build anything.
1. Can you show me a human reviewing the output before it goes live?
Not "we have quality control." Show me the process. Where does a person sit in the workflow? What do they check? If they can't answer this specifically, their output is going to your customers unreviewed. That's how brand-damaging content gets published at scale — and Air Canada learned this the hard way when their AI chatbot invented a bereavement fare policy that didn't exist, leaving the company legally liable for a promise no human ever approved.
2. Which model are you using — and why that one specifically?
Most buyers never think to ask this. They should. There are dozens of models available at very different price points, capability levels, and environmental costs. A company that defaults to the most powerful model for every task either doesn't know what they're doing or is padding margins. The right answer involves actual reasoning: "For this task we use X because Y. For this other task we use a smaller, more efficient model because it's sufficient." Vagueness here is a red flag — and a sign they're burning your budget and the planet's resources unnecessarily.
3. What happens when it gets it wrong?
It will get it wrong. Every model hallucinates. Every system has edge cases. Ask for a specific example of a failure and how they handled it. The answer tells you everything about whether they've thought seriously about what they're building — or whether they'll disappear when something breaks. If they say it never fails, leave.
4. How do you measure whether this is actually saving time?
"AI will save you hours every week" is easy to say. Measuring it is harder and most people don't bother. If they can't tell you how they'll baseline your current process, track improvement over time, and report back with real numbers — they're selling you a feeling, not a result. More than 80% of executives surveyed by McKinsey said generative AI has not yet moved enterprise-level earnings in any tangible way. The ones that do move the needle are the ones where someone defined what success looked like before they started.
5. Is there a human we can call when something breaks?
Not a chatbot. Not a ticketing system. A person. AI implementations break — models update, processes change, staff turn over and nobody remembers how it works. The companies worth working with know this and plan for it. The ones who disappear after go-live know it too. That's why they disappear.
Why this matters beyond your business
When buyers get burned by a tinfoil spaceship, they don't just stop trusting that vendor. They stop trusting the whole category. The estate agent who says "everyone's claiming they can do it and most of it's bullshit" isn't wrong — and that cynicism means businesses that genuinely need help aren't getting it, and companies doing it properly can't get in the room.
AI done right — the right model for the right task, humans accountable for every output, honest about where it helps and where it doesn't — is genuinely valuable. We've seen it save businesses real time and real money.
But you have to be able to find it first.
Ask the five questions. Listen carefully to the answers. The ones who can answer all five specifically, with evidence, without flinching — those are your spaceships.
Stop the Slop is a manifesto by 4142 Ltd. If your company builds with AI and holds itself to the same standard, sign it.