AI Is a Curve, Not a Point
- Roberto Massa
- 4 hours ago
- 5 min read
The Real Battle Will Be Managing Abundance
Much of the business conversation around artificial intelligence starts from a static snapshot: Which model is the most advanced? How many jobs will disappear? Which company is ahead? The United States or China? How much can we reduce costs this year?
These are legitimate questions, but they are not enough. They attempt to explain, through a fixed image, a phenomenon that is moving rapidly: model capabilities are changing, costs are falling, adoption is accelerating, and control over the infrastructure required to operate AI at scale is being reshaped.
Artificial intelligence is better understood as a curve than as a point. And looking at that curve forces us to ask a more strategic question:
What happens when capabilities that were scarce, expensive, and highly specialized yesterday begin to become accessible, replicable, and potentially abundant?
The answer matters because abundance does not eliminate scarcity. It shifts it.
Four Curves in Motion
The first curve is capability: Today’s systems can already perform tasks that, just a few years ago, required specialized expertise: interpreting complex documents, generating code, analyzing large volumes of information, identifying patterns, supporting research, modeling scenarios, and assisting operational decision-making.
The second is cost: A capability that today seems extraordinary, exclusive, or prohibitively expensive may soon become a standard feature embedded in a common enterprise application. The cost of reaching certain levels of performance has steadily declined, while smaller, more specialized models are beginning to solve tasks that once required much larger infrastructures.
The third curve is diffusion: Having access to AI is not the same as transforming an organization with AI. Experimental use has expanded rapidly, but the deep redesign of processes, operating models, and decision-making remains far less common. This gap explains much of the distance between “testing AI” and capturing real economic value.
The fourth curve is power: Behind the apparent democratization of AI, critical resources remain concentrated: advanced semiconductors, data centers, energy, cloud platforms, foundation models, large data repositories, and specialized talent. AI may become more accessible at the surface while, at the same time, becoming more concentrated in the infrastructure that makes it possible.
This tension will be one of the most relevant strategic issues of the coming years.
For centuries, much of specialized knowledge came at a high cost.
Analyzing thousands of contracts, reviewing millions of transactions, personalizing educational content, monitoring security alerts, translating documentation, developing software, or generating scenario analyses required many hours of highly skilled human work.
AI is beginning to reduce the marginal cost of some of these activities, opening the possibility for analysis, knowledge generation, and certain intellectual capabilities to become far more abundant than before.
But an important warning is necessary: technological abundance does not equal strategic abundance. When producing analysis becomes cheap, the value of knowing which analysis deserves attention increases.
When content generation becomes abundant, the value of credibility increases.
When agents are capable of executing actions, identity, authorization, and traceability become critical.
And as computing capacity expands, energy, infrastructure, and technological sovereignty gain greater geopolitical importance.
The equation can be summarized as follows: AI reduces some forms of scarcity while creating others.
Therefore, identifying which resources will stop being scarce—and which ones will become decisive—is more important than trying to predict which model will dominate next.
We also need to reconsider the tendency to interpret AI exclusively as a zero-sum game.
Under this logic, if one company wins, another loses; if a machine performs an activity, a person becomes unnecessary; if one country concentrates technological capabilities, all others are inevitably left behind. These are zero-sum conclusions.
There are markets, budgets, jobs, and geopolitical rivalries where this logic does apply. However, using it to explain the entire AI economy leads to overly simplistic decisions.
AI can create NON-ZERO-SUM scenarios when it expands the total value available. This does not mean denying competition, but rather recognizing that some innovations increase the size of the market, productivity, and our collective ability to solve problems.
Consider, for example, cybersecurity.
A SOC can use AI to classify enormous volumes of alerts, correlate signals, and prioritize potential threats. The weakest response would be to use that productivity solely to ask how many analysts can be eliminated. The strategically interesting question is different:
How much can we increase our actual protection capabilities with the same resources?
If AI reduces the time spent filtering noise, analysts can focus on investigating threats, validating anomalies, improving controls, and making risk-based decisions. The organization increases its productivity; professionals expand their capabilities; customers receive better protection; and the resilience of an entire value chain can improve.
That is non-zero-sum.
It does not mean everyone wins automatically. It means the total amount of value can grow. Technology expands the pie; business, regulatory, and institutional decisions influence how that value is distributed.
The most common mistake is using an exponential technology to optimize processes designed for another era.
Automating a report, accelerating the creation of a presentation, or introducing a copilot can generate real efficiencies. But there is a profound difference between automating an activity and redesigning an entire system of work.
Organizations that capture the greatest value from AI do not simply add another tool to existing workflows. They redesign processes, roles, decision-making mechanisms, data, controls, and customer experiences. They use AI to create new capabilities for growth and innovation—not merely to reduce costs.
This changes the conversation in the executive committee. The CEO should ask: Where does AI require us to change our operating model—or even our business model? The CFO should identify where the declining cost of intelligence changes the economics of a process, product, or service. And the CISO should assess how the risk surface evolves when people, machines, and autonomous agents can access information, make decisions, and execute actions.
The entire leadership team shares one responsibility: deciding which human capabilities are worth expanding before deciding which human costs can be eliminated.
A mature AI strategy should begin, at minimum, with five questions:
Which capability that is expensive today could become practically abundant within our business?
If that capability becomes cheaper, where will the new scarcity emerge: data, talent, energy, security, attention, integration, or trust?
Are we using AI to accelerate existing processes, or to redesign how we create value?
What dependencies are we accumulating across infrastructure, models, data, and vendors?
Does our governance enable us to scale with confidence, or does it merely attempt to contain risks after they emerge?
These questions remain relevant even if the leading model changes tomorrow. They place strategy above the tool and position governance as infrastructure.
Governance is often portrayed as a constraint on innovation. That interpretation is mistaken.
The abundance of digital capabilities also multiplies the potential for error, misuse, and attack. More agents mean more identities. More automation means more decisions made at machine speed. More data being processed creates greater exposure. More content makes it harder to verify provenance and authenticity.
That is why security, privacy, traceability, human oversight, and accountability are not peripheral compliance requirements. They are economic infrastructure. A company that cannot trust the outputs of its systems will struggle to automate consequential decisions. An organization that cannot govern the identities and permissions of its agents will struggle to grant them autonomy.
An ecosystem that cannot share information securely will limit its ability to collaborate, innovate, and respond to systemic risks.
Well-designed governance does not slow adoption. It makes adoption scalable.
What will happen to our organization when intelligence applied to specific tasks becomes inexpensive, ubiquitous, and available on demand?
That is where real transformation begins.
Leadership in AI will not consist solely of adopting the most powerful model or automating faster than competitors. It will consist of designing organizations capable of moving along these curves: capturing productivity, expanding human capabilities, managing new dependencies, protecting trust, and turning potential technological abundance into tangible economic and social value.
Technology will make abundance possible.
Strategy will determine who can turn it into value.
Roberto Massa
Director, Onistec Business Intelligence & Governance
Context References: Stanford Institute for Human-Centered AI, AI Index Report 2026; McKinsey & Company, The State of AI 2025; NIST, AI Risk Management Framework; Cloudflare Zero Trust; BeyondTrust; Identity First by Onistec.
Comments