At the bluegain CxO Luncheon during the World Economic Forum Annual Meeting in Davos 2026, one theme framed the room: what meaningful tomorrow can we shape with the courage to leave yesterday’s recipes behind? In the context of AI, that question is no longer philosophical – it is operational.

Leaders have a real window of opportunity right now. Expectations are rising, timelines are shrinking, and every board conversation seems to cover the same question: “Do we have the right AI strategy?”

AI is rapidly reducing the cost of conceptual work – text, analysis, reasoning, and increasingly software itself. As intelligence gets cheaper, advantage shifts from accessing intelligence to embedding it into operating systems. In this context, the more accurate question is quieter – and more structural: what happens to advantage when intelligence becomes cheap?

A basic economic unit is shifting. The marginal cost of producing conceptual work – text, analysis, structured reasoning, and increasingly software code – is collapsing relative to human labor. Token economics already reveals the direction of travel: intelligence is becoming abundant, price-segmented, and relentlessly cheaper at the margin. This is not just another productivity wave. It is a re-pricing of cognition itself. When the marginal cost of thinking declines, the structure of competition shifts with it.

When intelligence becomes abundant, the value moves away from “having the capability” and toward how you operationalize it: governance, workflow ownership, trust, security, and the cadence of execution. This is where NewLeadership shows what it is really capable of, namely implementing successful system renewal.

Commoditized magic: the Einstein moment where breakthroughs become baseline 

The defining paradox of the Intelligent Age is that what feels impossible becomes normal faster than organizations can adapt. This is the Einstein moment of commoditized magic: a state-of-the-art capability appears, and within a short diffusion cycle it becomes baseline. The window in which “being first” creates durable advantage is shrinking and as the window shrinks, strategy has to move up the stack.

The economic consequence follows quickly. When intelligence becomes cheaper and widely available, margins compress. Durability shifts away from the model itself toward the systems that operationalize it.

That forces a reset in familiar assumptions:

• “We have the best engineers” matters less when code generation becomes abundant.
• “We’re integrated” matters less when creation becomes modular and replaceable.
• “We have built a sticky business model ” matters less when switching costs for customers fall.

The moat therefore migrates. It sits less in the model and more in what surrounds it: domain constraints, workflow ownership, operating discipline, trust layers, and the ability to embed intelligence inside real processes. The winning companies will not be the ones who touch the newest magic first. They will be the ones who turn commoditized intelligence into repeatable outcomes.

What survives is compounding value:

• continuous improvement loops that get better with usage
• durable workflow ownership (so you are not a feature, you are the process)
• governance and security that enable adoption at scale
• monetization tied to outcomes, not activity
• feedback loops built on operational trust

For startups and investors, this weakens legacy licensing models. Paying for “work” begins replacing paying for “access.” For industrial corporates, the implication is deeper: slow decision cycles and human-heavy operating structures become strategic debt in a world where competitors’ cost curves collapse.

Software on Demand: From Products to Disposable Artifacts

One of the most under-discussed consequences of cheap intelligence is a redefinition of software itself.
Software doesn’t disappear. It becomes mass-customizable. The last mile moves to the user. Task-specific tools get generated on the spot, used, and discarded.

When software becomes ephemeral, SaaS-era assumptions weaken:

• lock-in becomes harder to sustain
• backward compatibility becomes less sacred
• integration costs become less of a prison sentence
• value moves to reliability, trust, governance, and outcomes

The metaphor that fits is “AI as Excel 2.0”: a platform layer that allows non-specialists to create functional software artifacts on demand. The strategic implication is clear: the product is no longer the artifact. The product becomes the system that makes artifacts safe, governed, and dependable inside the business. This shifts the buy/build question from features to governance and integration capability.

In other words, parts of software become disposable. Instead of maintaining a growing stack of durable applications, organizations increasingly generate “throw-away software” – artifacts created for a task, used briefly, and replaced by the next generation. The “40 cents” idea is not about a fixed price point. It’s about inevitability: as token costs fall, the cost of generating useful software artifacts trends toward negligible.

Tech Stack Layers: compute, deep-tech bottlenecks, and production reliability

As the application layer gets easier, defensibility doesn’t vanish. It migrates. Compute at scale remains a durable value layer, and policy institutions are already mapping the competitive dynamics of AI infrastructure. [OECD]

At the same time, not only giants win. Every deep tech stack contains bottlenecks – pockets of expertise where small teams with unusually strong specialists can create disproportionate advantage. Security is the canonical example. In those pockets, real engineering advantage still matters. Packaging doesn’t. Capital is also responding at infrastructure scale, and markets are increasingly sensitive to whether AI capex converts into growth rather than narrative.[ Reuters]

The pattern is consistent: as surface capability commoditizes, advantage concentrates in control points below it – compute, security, reliability engineering, and the operational plumbing that makes AI dependable in production.

Data Advantage: From Ownership to Permissioned Trust

Moats still exist. They move. Domain expertise can remain defensible in the near term where it reflects specialized knowledge and feedback loops that cannot be trivially reproduced. Proprietary data can still matter when it is truly unique and costly to replicate.

The longer arc tilts toward customer-owned data. Defensibility therefore depends less on captivity and more on permissioned access: trust-based exchange, consent, and the ability to generate compounding advantage from data without relying on hoarding. This is where governance becomes a growth enabler rather than a compliance tax. In an age of abundant intelligence, legitimacy/ trust becomes strategic – and governance becomes the mechanism that earns it.

Implications: What this means for industrial corporates

Physical assets can still be moats. They will not protect an outdated operating model. Even in asset-heavy sectors, the enterprise layer is where competitiveness shifts first: sales, service, planning, compliance, customer interaction, and decision cadence. Intelligence raises the ceiling and the baseline. Competitors become faster and cheaper – and increasingly, better.

The bottleneck is rarely tool availability. It is whether the organization can turn intelligence into repeatable execution without breaking trust. It is organizational adaptation: evaluation discipline, security posture, privacy readiness, data foundations, and execution rhythms that evolve weekly. Preparing takes time – and time is the one asset this transition punishes. The leadership task is not to “add AI.” It is to rewire the system so intelligence can be deployed safely, repeatedly, and measurably – without breaking trust.

Human plus-plus: Turning Abundance Into Advantage

The most constructive reading of this moment is not replacement. It is augmentation.

For CEOs, three strategic priorities follow from this shift.

• Build systems that compound intelligence.
  Organizations should focus less on acquiring the latest models and more on embedding intelligence into repeatable workflows that improve with usage.
• Invest in trust infrastructure.
  Governance, security, data permissions, and operational reliability become the foundations that allow AI capabilities to scale safely across the enterprise.
• Redesign work around augmentation.
  The real opportunity lies in combining human judgment with machine capability – allowing organizations to move faster while maintaining accountability and trust.

Cheap intelligence does not eliminate strategy. It raises the bar for it. The winning organizations will be those who turn commoditized capability into trusted execution – and make it compound over time.

This article is adapted from a keynote delivered by Yariv Adan at the bluegain CxO Luncheon during the World Economic Forum Annual Meeting in Davos (WEF 2026).

/ About the Speaker

• Yariv Adan is an AI pioneer, investor, and Managing Partner at Ellipsis VC, where he invests in and supports early-stage AI deep-tech companies. Previously, he spent 17 years at Google, rising to Senior Director of Product Management and helping shape some of the company’s most influential AI products. He is the co-founder of Google Assistant and Google Lens, and also led product management for Google Cloud Conversational AI and Applied Generative AI. He additionally served as site lead for Google’s Zurich office, overseeing one of the company’s largest engineering hubs. Alongside his product and investment work, he explores the intersection of AI, art, and creativity through initiatives such as AI-assisted art platforms, and his generative artwork “Rosy AI” was featured at the Swiss Pavilion during CES in Las Vegas. He is also passionate about AI for social impact and has contributed to global discussions on humanitarian applications of artificial intelligence, including through a United Nations panel on AI for social good.