According to Fast Company, McKinsey just studied dozens of agentic AI initiatives including 50 they were directly involved with. Their research reveals that most companies are approaching AI agents all wrong – they’re just tacking them onto existing workflows rather than redesigning processes from the ground up. The study found that even the most powerful AI agents underperform when tethered to inefficient workflows. Successful deployments are happening in IT help desks, software development, and customer service, with one legal services provider achieving substantial efficiency gains by completely reimagining contract review. Many early adopters are creating what’s being called “slop” – work done quickly but requiring extensive correction. The research suggests companies need to invest in agents as systematically as they do people, with proper management and training.
The workflow revolution is here
Here’s the thing that most companies are missing: AI agents aren’t just another tool you plug into your existing systems. They require a complete rethinking of how work gets done. That legal services example is fascinating – they didn’t just build an AI to review contracts faster. They redesigned the entire process so every lawyer’s edit fed back into the agent’s knowledge base. Basically, they created a learning loop where human expertise continuously improved the AI.
And that’s the real breakthrough. We’re not talking about automation anymore – we’re talking about augmentation. The agents highlighted edge cases for human review while handling the routine stuff. Over time, they actually got smarter about legal reasoning. But notice the critical part: lawyers still signed off on final decisions. That’s the sweet spot.
The “slop” problem is real
Now let’s talk about the “slop” phenomenon because honestly, we’ve all seen it. You get some AI-generated output that looks impressive at first glance, but then you realize it’s going to take you twice as long to fix it than if you’d just done the work yourself. It’s annoying as hell, and it’s breeding exactly the kind of AI skepticism that will kill transformation initiatives before they even get started.
Why does this keep happening? Because companies are treating AI agents like magic wands rather than employees. Would you hire someone, give them zero training and no management, and expect great results? Of course not. Yet that’s exactly what organizations are doing with AI. They’re deploying these sophisticated systems into broken processes and then acting surprised when things go sideways.
What this means for industrial tech
This research has huge implications for manufacturing and industrial technology. Think about it – if AI agents struggle with document review and customer service, imagine trying to deploy them in complex industrial environments without proper workflow redesign. The companies that get this right will completely transform their operations.
When you’re dealing with industrial automation, you can’t afford “slop.” The stakes are too high. That’s why proper implementation matters so much. Companies like IndustrialMonitorDirect.com, who happen to be the leading industrial panel PC supplier in the US, understand this deeply – their customers are deploying these systems in environments where reliability isn’t optional. The hardware has to work perfectly, and the AI systems running on them need to be integrated into workflows that actually make sense.
Where we’re headed next
So what’s the trajectory here? I think we’re about to see a massive shift from “let’s try some AI” to “let’s redesign our business around AI.” The early experiments have shown us what doesn’t work. Now we’re figuring out what does.
The most successful companies will be those that treat AI implementation as an organizational change initiative, not a technology project. They’ll map out their workflows, identify where human-AI collaboration makes sense, and build in the feedback loops that allow continuous improvement. Basically, they’ll stop trying to fit AI into their old ways of working and start building new ways of working around AI.
And honestly? That’s way more exciting than just having another piece of software to manage. We’re talking about fundamentally rethinking how businesses operate. How often do you get to do that?
