The hype cycle trap
Every few years, there's a technology that's supposed to save businesses from themselves. Before AI, it was the cloud. Before that, it was ERP. Before that, it was the internet. Each time, the same thing happens: companies with broken processes buy the new thing, bolt it onto the mess, and wonder why nothing got better.
I watch this happen with ERP all the time. A company with chaotic inventory buys a six-figure system and expects it to impose order. It doesn't. The system just gives them faster, more expensive chaos. Now AI is the new ERP in this story. Same trap, shinier packaging.
What AI actually does
Strip away the marketing and AI does three things: it recognizes patterns, it generates content, and it follows instructions really fast. That's it. It's not magic. It's not judgment. It's a very fast intern who never sleeps and never questions your instructions.
Give that intern a clean process and clear instructions, and they'll do amazing work. Give them a broken process and vague instructions, and they'll confidently produce garbage at a rate no human team could match. The speed is the danger. A human doing a broken process slowly gives you time to notice. AI doing it at scale means the damage is done before your coffee cools.
The broken process hall of fame
The approval chain nobody understands. Your purchase approvals go through four people, nobody knows why, and half the time they sit in someone's inbox for a week. Add AI to "streamline approvals" and congratulations: now the confusion moves faster. The AI routes things efficiently through a process that shouldn't exist. You didn't need automation. You needed to ask why four people approve a $200 purchase.
The data nobody trusts. Your CRM is full of duplicates, half the contacts are stale, and salespeople enter data "their own way." Now you want AI to score leads and predict revenue. The AI will happily do it. It'll give you beautiful dashboards built on garbage. Garbage in, gospel out, because the charts look so convincing that nobody questions them.
The meeting that should've been an email. Your team spends ten hours a week in status meetings. Someone suggests an AI meeting assistant to summarize them. Now you have perfect summaries of meetings that shouldn't happen. The AI didn't fix anything. It just documented the waste more efficiently.
Fix first, automate second
Here's the unsexy truth: 80% of the value people expect from AI comes from fixing the process, not from the AI itself. When you map out a workflow, remove the stupid steps, clarify who's responsible for what, and define what "done" looks like, you've already won. The AI is just the cherry on top.
I tell my ERP clients the same thing. Don't buy software to fix a process problem. Fix the process, then buy software that supports the fixed version. The companies that do it in that order succeed. The ones that reverse it end up with expensive shelfware and the same problems they started with.
What good looks like
A company with a clean lead follow-up process adds AI to draft personalized follow-up emails. The process was already working; AI just made it faster and more consistent. That's a win.
A company with clear inventory procedures uses AI to predict reorder points. The procedures were solid; AI added foresight. That's a win.
A company with a well-defined customer onboarding checklist uses AI to guide new customers through it conversationally. The checklist was already good; AI made it feel personal. That's a win.
See the pattern? In every case, the process came first. AI was the multiplier, not the foundation.
The question to ask before any AI project
Forget "what can AI do for us?" Ask this instead: "If we had to do this process with pen and paper, would it make sense?" If the answer is no, AI won't save it. If the answer is yes but it's slow, then AI might be exactly what you need.
Walk through your process step by step. For each step, ask: does this add value, or is it here because we've always done it this way? Cut the dead weight. Clarify the handoffs. Define the outputs. Then, and only then, ask where AI fits.
The speed trap
There's one more danger nobody talks about: AI makes bad decisions faster, but it also makes them harder to catch. When a human messes up, you see it. The email has a typo, the numbers don't add up, something feels off. When AI messes up at scale, the errors are systematic and invisible. Every output looks polished. Every number is formatted correctly. The wrongness hides behind professionalism.
This is why human oversight matters more with AI, not less. The faster the system, the more important the checkpoints. Build review steps into your AI workflows. Spot-check outputs. And never, ever let AI make decisions you wouldn't trust an intern to make unsupervised.
Bottom line
AI is a multiplier. Multiply a good process and you get greatness. Multiply a broken process and you get a faster disaster. The work isn't in the AI. The work is in the process. Do the boring work first. Map it, fix it, simplify it. Then let AI run it at speed.
The companies that understand this are quietly winning. The ones chasing the hype are quietly building faster ways to fail. Pick your camp.
FAQ
So should we just skip AI entirely?
No. AI is genuinely useful. Just don't expect it to fix what's broken. Use it to accelerate what already works, and fix the broken stuff the old-fashioned way: by thinking about it.
What's the first process we should fix before adding AI?
Whatever touches money or customers. Lead follow-up, invoicing, order fulfillment, customer support. These are high-impact and usually messy. Clean them up first and you'll see immediate returns, with or without AI.
How do we know if a process is "good enough" for AI?
Can you write it down as a clear sequence of steps? Does everyone follow the same steps? Are the outputs consistent? If yes, it's ready. If people do it differently every time, fix that first.
Isn't this just common sense?
Yes. That's what makes it so easy to ignore. Common sense doesn't sell software licenses or get conference keynotes. But it's what separates the companies that benefit from AI from the ones that just pay for it.
David Strausser is a business development executive specializing in ERP solutions for small and mid-sized businesses, including Odoo and SAP Business One. He helps companies fix their processes before they automate them, because he's seen what happens when they don't.
Tired of broken processes? Book a meeting and let's talk about fixing the foundation first.
