Hype Cycle for AI 2025

  • A grounded reading
  • October 7, 2025

Artificial Intelligence is at a pivotal moment. After years of experimentation, hype, and big promises, organizations are now facing a deeper question: How do we make this actually work in our operations?

Gartner’s 2025 Hype Cycle for AI offers a compass to navigate this moment. But beyond the chart and technical terms, at EDSA we believe it’s critical to bring these concepts down to earth—so that CIOs, tech leaders, and product owners can make informed, realistic, and context-aware decisions.

Gartner’s Hype Cycle is a visual tool developed by Gartner to represent the maturity, adoption, and commercial application of emerging technologies. It’s widely used to understand how technologies evolve from their emergence to widespread adoption—or disappearance.

logo Gartner

Two races, one choice: which are we in?

One of Gartner’s most valuable contributions to understanding AI’s current state is the distinction between two parallel races:

At EDSA, we work with companies that know they’re not in the game to "win the model race," but to achieve tangible results. This means working with focus, finding the right pace, and building internal capabilities. It’s not about running faster—it’s about running smarter.

Where are we now? The Trough of Disillusionment

According to Gartner, Generative AI has already entered the Trough of Disillusionment. This means that after the initial excitement, organizations are becoming aware of its limitations and challenges:

This doesn’t mean AI is failing—it means we’ve entered a stage that requires maturity, strategy, and execution.

Steady pace or fast track: define your AI strategy

Gartner proposes thinking of AI adoption as a race where every organization sets its own pace. Not all need to move at the same speed. Two valid approaches:

At EDSA, we help define the right pace based on your current capabilities, business vision, and data infrastructure. Faster is not always better—clarity and consistency matter more.

Expected AI outcomes: think beyond the hype

In the right race and at the right pace, what matters is concrete results. Gartner identifies three types:

Most organizations are still focused on employee productivity. But even there, the landscape is more complex than it seems.

For example, GenAI doesn’t impact all employees equally:

The challenge lies in identifying Deep Productivity Zones: the combinations of experience and complexity where AI can truly unlock value.

The best use cases typically combine either low experience + low complexity, or high experience + high complexity. At EDSA, we help organizations discover these strategic use cases.

Laying the groundwork for scalable AI

While GenAI cools off a bit on the hype graph, new methodologies and key practices are emerging to set the foundation for a scalable future:

  • AI Engineering: The set of practices required to build AI solutions that are secure, repeatable, and scalable—from pipeline integration to version control for models.
  • ModelOps: Governance of the full lifecycle of AI and advanced analytics models.
  • AI-ready data: Preparing, curating, and managing structured and unstructured data to feed real AI solutions.

The harsh reality: Over 50% of enterprise data isn’t AI-ready. In many cases, it’s messy, poorly labeled, or inaccessible. Worse still, data access permissions are often undefined. If not properly configured from the start, a model might access (and display) sensitive information.

From tech stack to tech sandwich

In this new landscape, the old idea of the “tech stack” falls short. Gartner introduces a more fitting metaphor: the tech sandwich.

The challenge is to bridge these two layers, build frictionless governance, and allow AI to emerge from everywhere—without ending up in a mess of broken APIs, shadow IT, and disjointed systems.

At EDSA, we’re already working with clients to design these tech sandwiches—helping define the right data, tools, and permissions at each layer, and ensuring they work together without blocking innovation.

Cost, PoCs, and realism: the new AI value management

One of the biggest risks today isn’t technical—it’s financial. Implementing AI can lead to 500% to 1000% budgeting errors if you don’t understand how costs scale: inference, data preparation, API calls, grounding, context handling, and more.

That’s why at EDSA, we recommend rethinking the PoC. It shouldn’t just prove the tech works—it must prove the value. You need to understand if what works in small scale can scale sustainably. It’s not about “seeing if it’s cool”—it’s about understanding how much it costs, how much it saves, and what it enables.

So what does this mean for you?

AI is no longer a trend—it’s a race. But not a sprint. It’s a race of direction. To move forward successfully, you need to:

At EDSA, that’s what we’re here for—to turn hype into real impact.

At EDSA, we’ll help you uncover your Deep Productivity Zones, assess your true data readiness, and build a concrete, actionable roadmap to scale AI with real impact.

Let’s connect