A new playbook for AI: Move fast, manage spend
The article also points to a familiar enterprise failure mode: scaling AI on top of unresolved data fragmentation and weak accountability. Leaders should treat data readiness, model-routing choices, and responsible-AI controls as part of the operating model, not just technical prerequisites. That means assigning decision rights across architecture, product, finance, procurement, and risk teams so that no single function optimizes cost, speed, or compliance in isolation.
Perhaps the most important implication is portfolio discipline. AI initiatives should be funded and reviewed like continuously managed products, with outcome metrics that extend beyond usage or model performance. Adoption, workflow redesign, and benefits tracking matter as much as technical deployment. If users revert to legacy ways of working, even efficient AI can become waste.
Practical questions for leadership:
- Which AI use cases have defined cost-per-task and value-per-task measures?
- Where do data quality or integration gaps create hidden production risk?
- Who owns model selection, routing policy, and vendor-spend accountability?
- What adoption and business outcome metrics will determine whether to scale, redesign, or stop an initiative?
AI-native companies are growing up to 40% faster than traditional SaaS companies, according to Gartner. But this growth is colliding with a cost crisis.
One company, for instance, was billed half a billion dollars in a single month for AI usage. While it might have been an outlier, it was also an early warning. As enterprises race to become AI-native, a cost crisis collides with a performance mandate. Most leaders don’t have the visibility to manage either.
The pressure to modernize is real, but the shift from flat subscriptions to usage-based pricing has introduced a volatility for which companies aren’t built. Although the overall cost of models is coming down, the actual cost of completing a task with AI is, paradoxically, often higher than companies expect.
There are countless factors that can cause enterprise AI bills to soar with consumption-based usage contracts. Placing a spending cap on AI activity without accounting for the task’s purpose will also cap innovation. Being an AI-native enterprise, in my view, comes down to orchestrating three disciplines: an intelligent foundation, intelligent economics and continuous optimization.
Sustainable AI advantage doesn’t come from deploying the most advanced models. Rather, it comes from getting these three pillars working together to deliver measurable business outcomes at enterprise scale. Here’s what that looks like in practice.
The groundwork most AI projects skip
At a glance, an AI-native company is one that built its operating model and workflows around AI from day one, which is far more doable for a startup than it is for a large enterprise running thousands of employees across dozens of systems.
That is where most enterprise modernization efforts stall. Intelligent foundations come first. Those are the unglamorous data platforms and responsible AI guardrails. If you skip building those, every other discipline inherits the same blind spots downstream.
The scale of risk is easy to underestimate. Through 2026, an estimated 60% of AI projects will be abandoned simply because the underlying data was never AI-ready, according to Gartner. The pattern is familiar: A model performs well in a demo because someone spent a week quietly cleaning the data set behind the scenes, but then falls apart the moment it hits production.
That’s rarely a model problem. It’s a foundation problem: fragmented systems that were never reconciled. Data modernization isn’t a preliminary step to get through before the “real” AI work begins; it’s the infrastructure upon which everything else in the AI-native enterprise sits.
Enterprises that treat data readiness as an ongoing discipline, not a one-time cleanup, are the ones with a foundation solid enough to support the other pillars.
Look beyond usage costs with intelligent economics
But being AI-native isn’t just about applying technology indiscriminately, nor is it about placing arbitrary spending limits on innovation.
In reality, some use cases will drive up consumption costs far faster than others, and high usage can absolutely be justified when it creates real value. Meanwhile, data shows the most productive engineers are also the ones consuming the most tokens: High spend and high value often move together.
The job for leadership in this scenario isn’t to ration AI, but to govern it. Here, intelligent economics treats every AI-driven action as a unit of cost and value, weighing what a task costs to run against the business outcome it unlocks.
To do this, enterprises need to access the right data. According to a recent KPMG report, only 26% of organizations have real-time visibility into the cost of running AI at scale.
Enterprises that master how to measure usage and performance can then work with inference economics to manage initiatives with rigor. Inference economics refers to the recurring operational cost of every task or query handed to AI, and managing it requires rigor. The payoff for getting this right is concrete. Legal AI company Harvey, for one, recently tested a hybrid setup that outperformed running the frontier model on everything — at 39% of the cost.
The lesson generalizes: the most expensive model isn’t always the most accurate for a given task, and routing intelligently can beat brute force on both quality and price.
The road to continuous optimization
Cost visibility tells only half the story. Enterprises also need to know whether AI is moving the metrics that matter, and that question doesn’t go away once a workflow ships. It just changes shape.
Such is the discipline of continuous optimization: Treating AI systems less like a project with a finish line and more like infrastructure that has to be watched, tuned and adjusted as conditions change. In practice, that means anchoring AI initiatives to outcomes leadership actually cares about.
It means putting as much energy into adoption as into deployment; the best-governed AI initiative delivers little value if the people meant to use it revert to old workflows within a few months.
Continuous optimization thus closes the loop between the other disciplines. It’s how enterprises know whether their foundation and cost governance are actually compounding into value.
The journey to AI-native enterprise
In 2026, business value from AI is the sum of intelligent workflows, cost efficiency and adoption.
The most competitive advantages will be associated with AI-native enterprises built around an intelligent foundation, intelligent economics and continuous optimization, with the ability to orchestrate innovation to deliver business growth objectives while keeping costs sustainable.
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.