Why Generative AI Isn’t Working for Most Companies (Yet)

Why Generative AI Isn’t Working for Most Companies (Yet)

AI has moved from hype to headlines—and now, to budgets. Yet when I speak with CX leaders across industries, one message keeps coming through: we’re still figuring out how to make it actually work.

The truth is, while generative AI has already reshaped how we communicate, write, and search, the path to enterprise impact is far less linear. Most organizations are still navigating early pilots, disconnected tools, and proof-of-concepts that haven’t yet scaled.

Why?

Because deploying AI in a business context—especially in high-volume, high-urgency environments like CX—isn’t just a technical challenge. It’s a systems challenge.

The Hidden Complexity of AI in Real Business Settings

Consumer tools like ChatGPT feel seamless because they’re trained on enormous public datasets and used in controlled contexts. But business environments are different.

Internal data is fragmented, knowledge is tribal, and language is full of nuance—like product nicknames, acronyms, and workflows that don’t exist outside the organization. AI can’t just “plug in” and perform. It needs context. It needs structure. It needs to learn the way your people work.

Many organizations try to solve this with knowledge retrieval techniques. But if the knowledge isn’t clean, current, and connected, the system spends more time searching than serving customer or team members.

I’ve said before that using RAG is like taking an open-book test with a giant, messy textbook. If you don’t know where to look, it’s slow...and you might not find the right answer.

The Real Cost of AI Isn't Just Technical

A lot of the conversation around AI cost focuses on models, infrastructure, or vendor pricing. But the real cost is often hidden in the prep work: cleaning up knowledge bases, tagging documentation, and mapping processes across siloed systems.

Until that foundation is in place, even the most powerful models struggle to deliver meaningful outcomes. This is where many enterprise AI projects stall—not because of lack of ambition, but because of messy reality.

From Pilots to Performance: What AI Success Looks Like

What separates teams that scale AI from those that get stuck in pilot purgatory?

It’s not just better models or more funding—it’s clarity on outcomes.

When AI is embedded into workflows—not layered on top—organizations start seeing measurable gains:

  • Faster resolution times and better SLA adherence
  • Smaller backlogs and more efficient triage
  • Reduced time-to-proficiency for new hires
  • Higher output per agent or rep, without adding headcount
  • Increased deflection through intelligent self-service

These are the metrics that move decision-makers. And they only happen when AI is grounded in operational reality.

Designing for Systems, Not Just Use Cases

One of the most common mistakes I see companies make is treating AI as a tool for isolated use cases—enterprise search, summarization, ticket triage, etc. But real impact comes from designing systems where AI plays a central, continuous role.

That means rethinking workflows. Reducing tool sprawl. Making it easier for employees to access what they need, when they need it—with context built in.

It’s not about bolting on AI to what you already have. It’s about designing around it.

What Comes Next

Looking ahead, we’ll likely see several major shifts in how organizations adopt and operationalize AI:

  • Consolidation of fragmented workflows into AI-native systems.
  • Decline of legacy, static knowledge tools in favor of dynamic, context-aware platforms.
  • Increased pressure to prove AI ROI with real business metrics—not just productivity anecdotes.

Article content

Ultimately, this isn’t about following the hype curve. It’s about building the infrastructure for long-term adaptability across complex enterprises.

AI won’t revolutionize customer experience overnight, but with thoughtful design and a focus on measurable value, it can become one of the most powerful enablers of modern CX.

The future isn’t about experimenting with AI. It’s about building with it.

Shimon Liberman

Head of Developer Success & Support at Frontegg

2mo

Love it! A real reflection of what’s the challenges out there. I only have one comment but I’ll tell you in person on Tuesday ;)

Patrick van Donselaar

Vice President, Shaping the future of enterprise cloud customer engagement through AI-driven innovation, scalable design, and automation strategy.

2mo

Insightful article! I see this a lot as well: organizations expect breakthrough results from (Gen)AI while treating it like a side project. But AI isn’t a plugin, it’s a shift in how work gets done. It demands strategic priority, intentional design, executive backing, and dedicated ownership. Without that, you’re just asking overstretched teams to layer disparate intelligence on top of dysfunction. In my opinion, if we expect systems-level impact, we have to intentionally build for it; technically, organizationally, and culturally.

To view or add a comment, sign in

Others also viewed

Explore topics