AI in the Enterprise: Harder Than the Hype
The hard part about AI isn’t the tech—it’s the humans
Like many of you, we’ve been working hard at our company to replace tasks that have been done manually for decades with AI solutions.
And honestly—it’s not all it’s cracked up to be. The paradox of AI as our efficiency savior was illustrated in last week’s hype cycle about MIT’s study that 95% of AI pilots in enterprises fail.
It’s not so much a problem with the technology. It’s with the implementation.
When it comes to bringing AI into a real operating environment, here are five challenges I’ve been running into.
1. Too Much Vaporware
Vendors are pitching “AI-powered” features at every turn. But they often don't live up to the hype—at least not yet. The solutions we’ve evaluated have some great features, but rarely enough to justify the cost and the massive change-management effort required.
2. The Demo Problem
It’s not enough to have cool AI features—you need sales teams who can actually show them off. I literally sat through a demo the other week where the vendor’s “AI” demo kicked off by showing my team how to generate reports using dropdown fields. 🤯
The AI is there. But the sales teams haven’t caught up.
3. AI Can’t Fix Your Bad Data
AI only works if it’s fed a diet of high-quality data. It’s not smart enough (yet) to figure out if your data is incomplete, inconsistent, or siloed. We can feed our entire pricing structure into an AI tool to help us optimize our pricing, but if our actual price and cost data isn’t accurate or clearly structured, we’ll be driving the wrong changes led by flawed AI insights.
Fragmented, ungoverned data is foundational dysfunction. Some scale companies can’t answer simple questions like "How many unique customers are in our Salesforce instance?" That chaos gets dragged into AI models, which then amplify confusion rather than create business ready insight.
4. This Is Where I Leave You
The tech may be cutting edge, but the people equation is still critical. Fear of job displacement is real and creates resistance among teams to adopt AI based tools. “If I don’t use it, you won’t know if it’s better than me!”
5. The Skills Gap
The stark reality is that many employees lack the technical foundation to use AI tools. It’s like handing someone a Formula 1 car when they’ve only been riding a scooter. Despite all the hype about “just writing a simple prompt,” most AI-enabled solutions today come with new (and complicated) interfaces, new workflows, and new ways of thinking that are designed for technologists and developers more than customer service and sales.
It’s hard to find the ROI in AI.
Put all of that together, and it’s no wonder many companies struggle to find a real return on their AI investments. AI is positioned as a magic bullet for productivity by AI advocates who haven’t actually tried to implement today’s solutions in today’s operating enterprises.
I still believe the future belongs to AI-first companies. Which means we can’t sit this one out—we just need to be smarter about how we implement it.
Tackling the Challenges
What are we doing to overcome these challenges? Gotta admit – it’s not easy. But here are three things I’m working on.
1. Prioritize AI-first startups over “AI-washed” incumbents. We’re leaning toward working with startups that are AI-first over established vendors with bolt-on AI features. When the technology isn’t an add-on but the foundation, the solutions tend to be stronger and more intuitive.
2. Invest in data first. Our very first AI initiative isn’t about AI at all—it’s about cleaning up our data. It’s not a fun job. But if the foundational data is unsound, AI can’t deliver meaningful results.
3. Start small with your most tech-savvy employees. Instead of rolling out an AI initiative across the whole company, we’re launching it for use with our most tech-savvy employees first. That helps us see where AI genuinely moves the needle without the noise of adoption challenges.
The Road Ahead
Eventually, there will be only two kinds of companies: those that use AI effectively, and those that are irrelevant. But eventually isn’t today.
Right now, the path to ROI with AI is messy. It requires remembering that people—not algorithms—are at the heart of every enterprise transformation.
I’d love to hear from others wrestling with this. Where do you see AI changing your operational scale? And what roadblocks are you running into on the way?
