
Last week we hosted our most recent Mayfield CXO Insight Call: “Beyond the Hype – Where is AI Gaining Traction?,” where Michael Chui, Senior Fellow at QuantumBlack, AI by McKinsey and Deepak Seth, Director at Gartner, covered both where we’re at in the hype cycle, and what’s beginning to emerge from it, as well as some real-world AI applications and strategies from industry pioneers.
Here were some of the key learnings:
Michael Chui, Senior Fellow at QuantumBlack, AI by McKinsey
Ever since ChatGPT became part of the global lexicon, there’s been a huge, renewed focus on AI, more specifically, generative AI. It’s important however, we don’t lose sight of the fact that there’s still tremendous room for analytical AI, for example, applying machine learning to forecasting, product recommendations, or conditional maintenance. There’s still over $10T worth of potential in that market alone. But now with generative AI, the door has been opened to a whole new slate of ideas, particularly for non-technical executives.
QuantumBlack has taken the time to map out where there’s potential value across the global economy, including over 60 separate use case families across many different sectors and functions. This includes areas where generative AI is creating most of the value, vs. cases where other kinds of AI, analytics, or even physical robotics are coming into play.
Looking at pure generative AI, it’s still relatively concentrated today in terms of where companies are seeing the most impact. Some of the areas seeing the most engagement are:
Beyond these top four early leaders, there’s plenty more fodder to look at across HR, supply chain, risk and legal, etc. The greatest margin and growth opportunities will vary widely depending on industry and function. As a consumer company, sales and marketing may be the most valuable, for a chemical company, their specialty may be within innovation and R&D. There has also been some movement in robotics and industrials – foundation models are being developed for robots. The providers of physical robotics are using AI to expand the breadth of their applications in the same way that agentic AI for knowledge work enables companies to deal with the long tail of what they’d otherwise have to program as exceptions. The goal is to enable robots to do things that don’t need to be programmed explicitly across all six axes. The promise is there, it’s just more important when dealing with the physical world to get things right the first time.
Today, we’re seeing large organizations engage with AI in three major ways:
Very few companies are in the position of being a “maker” today, although some truly are, with pharmaceuticals being a leading example. Then there’s a bit of a mix going on between the other two, which is very context dependent. There are many applications today where you can easily buy a solution off the shelf, and there are other cases where hooking something up to a vector database can be very effective. I think many companies are hoping that AI just gets incorporated as a base feature into the software they’re already using today.
In terms of pricing, some envision AI holding onto the traditional seats model, while others are thinking about charging per agent, or perhaps even getting on the same side of the table as their customer and charging for some percentage of value created. One of the challenges right now is that the cost to serve for GenAI is pretty high. Everyone wants to go to heaven, but no one wants to die in the process.
If you look at the companies where AI is really making an impact on their bottom line – today that number is vanishingly small. That doesn’t mean that the potential isn’t there, it just means it’s not there yet. Part of this is that businesses today haven’t yet reached a sufficient level of aspiration. If you only attack basic use cases with AI (since they are the low hanging fruit), it’s not necessarily enough to move the needle. You need to be looking at domains, functions, or end-to-end workflows. Unless teams attack an entire operation at once, they are going to undershoot in terms of value creation. Further, if there were AI-enabled agentic systems that could do multi-step reasoning, you would also see a much greater impact on the bottom line than the pre-agentic era we find ourselves in.
Today, the biggest potential for AI is in disrupting stagnant headcount and business models, but while technology is very cool and interesting, there’s a lot to business beyond it. Large successful firms also stay successful on account of things like client relationships, and their ability to adapt. Many of the outsourcing firms aren’t ignorant to the fact that AI has the potential to do what they do, at a lower cost, so you’ll see them moving in on this as well. There are many obvious predictions today when looking at AI’s potential, the question will be not only whether or not the technology makes it there. The laws of real business still apply.
Bonus Content
Deepak Seth, Director Analyst, Gartner
In the rapidly evolving world of technology, we often misjudge its immediate impact while failing to grasp its long-term potential. This is the essence of the hype cycle: technologies are initially overhyped, but only some fulfill their promise over time. An idea emerges, is spotlighted by the media, heralded as a transformative force, and then falters in delivering instant results, plunging into the trough of disillusionment. However, as understanding deepens, new trends solidify, and productivity gains materialize, the true value of technology becomes evident. The hype cycle is not about the technology itself but the surrounding excitement, which can overshadow the real benefits still being delivered even as hype wanes.
Take, for instance, Microsoft Excel, which recently celebrated its 40th anniversary. Despite having reached its productivity plateau and entering the trough of disillusionment decades ago, Excel continues to provide immense value. The lesson here is crucial: businesses should consider investing in technology not at the peak of its hype but when it begins its descent into the trough, where real value often lies.
Currently, generative AI stands at a pivotal point. According to Gartner, it has surpassed the peak of inflated expectations and is now descending the curve. Vendors face a significant gap between costs and revenues, with customers struggling to extract sufficient value to bridge this divide. A major hurdle is the technical implementation, with companies excelling in proof of concepts (POCs) but faltering when scaling these technologies enterprise-wide. This transition from the AI hype cycle to the AI ecosystem is a natural progression.
In today’s industry landscape, AI is ubiquitous, arriving from various directions. Previously, AI within companies was built and blended, but now it is embedded in software, requiring businesses to integrate their own AI solutions. This complexity, dubbed the “AI technology sandwich” by Gartner, involves 65% of AI entering through existing tools and only 35% being internally developed. The mix of centralized and decentralized data, built AI, blended AI, and embedded AI in software offerings creates a new tool stack within the AI ecosystem. Trust, risk, and security management layers are crucial, with AI committees and communities of practice oversight teams playing vital roles.
When a group within an organization seeks to introduce its own AI, a central committee should first approve it, followed by oversight from a TRiSM (Trust, Risk, Security Management) team for governance, trust, risk, and security implications. Integrating internal and external data adds further complexity, highlighting the critical role of data today in supporting AI implementations. Centers of excellence, initially housing PhDs, often fail by focusing solely on AI and neglecting classic IT involvement, which is essential for embedding capabilities into existing applications and streamlining deployment in production environments. Recognizing this gap is vital for achieving scalable AI success.
Moreover, many companies rush into defining use cases and POCs without establishing a comprehensive AI strategy. Prioritizing capabilities is key; starting with achievable back-office wins may be more feasible than attempting a complete organizational overhaul. Defining ambition is crucial before prioritizing use cases, as once ambition is clear, identifying them becomes more straightforward, guided by feasibility and ROI. If all you can afford is an inexpensive car, you shouldn’t waste energy trying to figure out how to build a race track for an expensive racecar.
At the board level, a disconnect persists between CEOs and CIOs regarding AI rollouts. CEOs, influenced by AI hype and pressured by boards, often clash with CIOs (and other technology leaders) who must balance risk and opportunity. This dynamic can lead to CEOs perceiving CIOs as obstructing AI efforts. The lack of IT representation on boards exacerbates this issue, with external players like Gartner sometimes needed to effectively communicate the CIO’s perspective to the CEO and board.
In summary, understanding the hype cycle’s nuances and navigating the intricate AI ecosystem are essential for harnessing technology’s true potential. By strategically timing investments, prioritizing capabilities, and fostering collaboration between CEOs and CIOs, businesses can unlock the transformative power of AI and drive sustainable success.