This article was authored by Drew Naukam, CEO of Gorilla Logic
Every week there’s another headline about AI agents coming for our jobs. But if you’re actually trying to build one of these things, you know the truth: it’s brutally hard work.
Getting the workflows right, preparing data, designing the agent behavior—it all takes massive effort, and the outcomes are unpredictable. You can’t build agents the way you build traditional software. Success means running experiments, measuring outputs, iterating constantly. That’s why most AI projects aren’t generating ROI yet.
We’re at the very beginning of this.
The Reality Behind the Hype
Does that mean AI is overhyped? No.
But the truth sits somewhere in the middle of the panic and the promises. What we’re seeing isn’t a failed technology—it’s a transformation that’s still being built. And there’s a useful parallel here. Thirty years ago, another enterprise technology wave forced companies to completely rethink how work got done.
When ERP Changed Everything
In the 1990s and early 2000s, companies spent billions on ERP programs meant to standardize “order-to-cash” and “procure-to-pay” processes. They bought SAP and Oracle, reengineered operations, launched huge change-management programs. The goal was to standardize everything.
And yes, if you manually processed accounts payable back then, your job probably changed. Or disappeared.
Standardization was the whole point. The technology was mature enough to enforce process discipline, and companies succeeded by aligning to those standards. ERP systems didn’t just automate; they codified how business worked.
From Standardization to Cognition
Now we’re doing it again. Companies are trying to automate and optimize workflows, but this time through AI agents.
The difference? There’s no equivalent of SAP for AI agents. No mature platform with proven blueprints. So companies are building these things themselves—defining processes, creating datasets, training models, testing and retesting. It’s slow, messy, and often frustrating for teams used to deterministic engineering. But this is what innovation looks like before it becomes infrastructure.
ERP changed how companies handled transactions. AI agents will change how they handle judgment, analysis, and decision-making. We’re still in the assembly stage — and that’s okay.
What Technology Leaders Should Do Now
For CIOs and CTOs, this moment demands realistic thinking. AI won’t follow a straight line, but you can still position your organization strategically.
Educate your board and executives.
Talk honestly about timelines and effort. This isn’t magic—it’s hard work and lots of iteration. Returns come from sustained effort, not pilot projects.
Empower grassroots experimentation.
Let your teams experiment with AI tools in their actual work. Celebrate small wins. Make it safe to fail, because that’s how people build real expertise.
Invest in process understanding.
Before you automate anything, map out the workflow. Figure out where human judgment matters and where an agent can actually help.
Build a culture of AI fluency.
Just like ERP demanded new process management skills, AI demands new capabilities: understanding data quality, knowing how to prompt models, validating outputs. That can’t live in one team. It needs to spread.
You might not be ready to deploy complex agents today. But you can absolutely create an environment where your teams are learning, experimenting, and getting ready.
The Hard Truth — and the Real Opportunity
Yes, AI is real. The promise is real. And yes, the reality on the ground is hard. But hard doesn’t mean it’s not working. It means we’re in the difficult early phase.
The companies that figure this out will be the ones treating AI as a long-term capability, not a short-term project. They’ll use this time to build expertise, establish discipline around data, and learn how to iterate effectively.
ERP reshaped business execution thirty years ago by codifying process. AI agents will do it again, but this time by augmenting cognition instead of just automating steps. The leaders who understand that difference—and prepare accordingly—will define what enterprise success looks like in the next decade.
