Everybody used to build out three-year strategy plans, because that was the typical industry flow of how things moved. That doesn’t matter anymore: enterprises now have to shift at the speed of AI.
Uber’s CTO made headlines earlier this year after revealing that the company spent its but because adoption exceeded every target.
More than halfway through 2026, it’s clear that Uber wasn’t an outlier, but more of an early warning.
Organizations are now confronting the reality that AI consumption can grow far faster than traditional technology budgets were designed to support. Enterprises that encouraged broad AI adoption are now introducing usage controls and governance mechanisms as token costs accumulate faster than anticipated.
The budget question is just one example of what’s happening on a broader level: enterprise infrastructure and operating models are often playing catch-up, too. Because AI isn’t just driving more volume; it is creating entirely new revenue and growth opportunities for your integration foundation. APIs become monetizable products, file transfers underpin higher-value transactions, B2B flows open new partner ecosystems.
Meanwhile, all of that growth doesn’t happen in a vacuum. It’s building sustained pressure on cloud or hybrid infrastructure that now has to absorb unpredictable, real-time demand. Enterprise systems were never designed for this kind of intensity.
Cloud taught us important lessons about scaling and growth
We’ve seen this movie before. In its early days, cloud was all about potential. Everybody was excited about the ability to scale seamlessly with high availability and reliability. Hyperscalers handed out compute coupons at every trade conference. And enterprise leaders quickly realized it could also drive spiraling costs.
It’s why we saw some significant repatriation from the cloud, or a rush to build private clouds instead of spending on hyperscalers.
What that era revealed is something that enterprise leaders are still internalizing today: cloud gives you elasticity, but real scalability comes from making deliberate architectural and business choices.
Cloud gave enterprises a powerful engine for scale, with the ability to launch new services and reach new markets faster than ever. But the organizations that captured that growth best didn’t just “lift-and-shift”. They rethought how their systems should operate, sometimes rearchitecting a process entirely to fit new realities.
Realistic cloud strategies evolved toward a more balanced hybrid model, where workloads can move across environments: scaling in the cloud when needed, staying anchored on-premises (or somewhere in between) where control, cost, or regulation requires it.
Something very similar is happening with AI.
AI is an acceleration engine that must be fueled and sustained
Leading AI labs are giving enterprises vast, cheap access to integrate AI into their workflows. But the economics are already changing: as AI providers search for more sustainable pricing models, organizations are largely shifting from “tokenmaxxing” to maximizing AI efficiency and chasing proof of ROI.
Enterprises are smart. They learned their lessons in the cloud era. But AI has created an acceleration of everything.
Leaders should now be asking themselves: does your infrastructure and the way you do things today support that future? Because today, you might handle 200 million API transactions a month. Once agents come into the mix, that could be 60 billion, and it could happen in two months versus three years.
The challenge isn’t simply processing more requests. Every API call or model invocation comes with operational and financial consequences. As the earlier examples of untethered AI adoption show, organizations are discovering that usage can grow far faster than budgets or governance models anticipated.
Organizations need visibility into how AI services are being consumed, which workloads are generating value, and where costs are accumulating before they become unsustainable.
This is where AI gateways are increasingly becoming critical. Just as API gateways helped enterprises govern, secure, and scale APIs, AI gateways provide centralized control over AI traffic and assets, helping organizations monitor usage, apply policies, optimize model selection, and manage costs as adoption grows.
But the hardest part is not so much building systems as it is running them at scale, sustainably. Beyond cost control, effectively governing AI consumption helps organizations build the confidence to expand AI into new products, services, and customer experiences without fear of runaway spending or operational instability.
Growth reveals what your systems can actually handle
Software is genuinely easier to build now. You can create a digital product, a website, an e-commerce front end, hook up all the payment systems, and do all of it within seconds by talking to a copilot.
But if everybody does that, you’re going to overwhelm the backend processing systems.
So, the question is no longer: can my technology stack scale with that demand? But rather: are we up to the challenge of turning that pressure into growth?
Organizations that can do both turn APIs into products, as opposed to treating them as mere technical interfaces. They onboard partners faster and open up new AI-enabled ecosystems. They launch new digital services without rebuilding their foundation each time.
Your tech stack should create growth, not just support it.
Take WeLab Bank, a digital-native bank operating at high scale across Asia. Their platform isn’t just built to handle volume; it enables them to continuously launch new financial services and expand their ecosystem. The ability to scale securely and reliably is what allows them to move faster in-market and capture open banking opportunities ahead of their competitors.
This is the shift: scaling becomes a competitive advantage when it lets you monetize, launch services, or onboard partners faster than anybody else.
What scaling for growth looks like in today’s enterprises
You can see this play out across different organizations and industries.
Take the pharmaceutical supply chain, for example: onboarding a new partner is a process that can take weeks. Imagine winning a new hospital network because you can onboard and transact with them in days. Meanwhile, your competitors are still buried in compliance paperwork.
And as demand signals, sourcing decisions, and partner discovery are increasingly driven by AI, the ability to execute compliant transactions instantly becomes a direct growth lever.
Cencora: the backbone that has to scale continuously
People love to talk about APIs and AI. But EDI is how the entire planet operates commercially. You cannot get an Amazon order without an EDI process in the backend: the warehouse would never get a shipment notification! For the world to function, that needs to stay in place.
Cencora is a perfect illustration of sustaining systems that already operate at massive scale.
Ranked #11 on the Fortune 500, this global pharmaceutical solutions company exchanges over 800,000 EDI messages a day with more than 23,000 trading partners, processing over $140 billion in annual revenue flows.
As their EDI technology lead Scott Marshall puts it: “Even a few minutes of unplanned EDI downtime could lead to millions of dollars in costs.” They’re growing EDI volume at roughly 10% year over year, and they’re already exploring containerized infrastructure and a move to cloud.
That’s the reality of backbone technology: the demand never stops growing. What’s more, reliable, high-volume B2B exchange isn’t just operational. The 10% YoY volume growth represents expanding business relationships, such as through onboarding new trading partners, not just “bigger pipes.”
AI doesn’t replace these workloads. Those transactions still have to move, and your business still has to operate.
Commerzbank: designing for the next order of magnitude
Commerzbank is a strong example of what happens when scalability is deliberately tied to growth outcomes, not just system performance.
A leading German bank with an international presence, their API usage grew from around 2 million requests per month in the first year on the platform to over 130 million just a few years later.
Today, Commerzbank process 6 billion API transactions a month. They went from 150 strategic APIs to planning for over 300, and from batch-based data exchange to near-real-time, monetized API services.
That’s exactly the kind of nonlinear growth curve that AI is about to steepen even further. And Commerzbank knows it. Rather than waiting to hit a wall, they’ve already started an initiative in Google Cloud to prepare for agentic capabilities, architecting for 10 times the scale of what they do today.
At the same time, they recognize that not everything moves to cloud. Some workloads need to remain on-premises, and those need to evolve too – that’s the hybrid reality.
Critically, Commerzbank didn’t just scale API transaction volumes. They used that platform to launch monetized API services and expand into fintech partnerships. The scale enabled the growth.
Commerzbank is a case study in getting ahead of it: investing in the operational foundation now, before the next wave of AI-driven demand makes it non-negotiable. And they tied that scale to business outcomes, such as cutting development effort by up to 80%, to build customer-facing services faster.
The sprawl problem that keeps compounding
Across these examples and conversations we have with customers every day, it’s clear that scaling doesn’t stay contained. It spreads across systems, teams, and technologies. Then, sprawl makes everything harder.
We used to talk about FTP sprawl, where everybody was doing file transfers around the organization (and bypassing secure channels for convenience – please don’t do that!). Then there was API sprawl. Now you have the sprawl of AI capabilities everywhere. Some are calling it “shadow AI.”
And here’s the thing: the minute data gets out into the AI ether, there is no coming back. Agents can spin up APIs and MCPs instantly. If you let them write and you don’t have a way to govern them, you’re going to have a million MCPs in your ecosystem overnight.
See also: Secure and Scalable Agentic AI: A Guide for Enterprise Leaders
Once again, if this is sounding familiar, it should be. With API sprawl, teams built and deployed APIs faster than organizations could discover, govern, or secure them. Our response was federated API management: creating visibility and governance across distributed APIs without forcing everything into a single centralized system.
Now, organizations are rapidly accumulating models, agents, MCP servers, and AI-powered services across business units and tech environments. We may be entering the era of federated AI management.
But here’s the opportunity that governed scaling can unlock: every API, every MCP server that’s properly registered and managed doesn’t just reduce risk. It becomes a reusable, potentially monetizable capability. And the same visibility that reduces governance risk can also help organizations understand the true cost and value of AI-powered services as they scale.
The gap between growth and readiness
Here’s what I keep coming back to: most organizations haven’t done the foundational work.
They haven’t modernized at scale the way they thought. What was good enough was good enough. If your workloads worked and it was cheap to run them, why change them?
Well, no one expected 10x capacity requirements almost overnight. And we didn’t expect it to happen in months instead of years!
The foundation work can no longer be deferred. Without it, organizations are just putting up a facade that will quickly be overwhelmed. Meanwhile, scalability also depends on flexibility: the ability to place and move workloads across environments, scaling in the cloud when needed, and remaining where control, regulation, or cost demands it.
That’s the challenge of AI: it scales so fast that enterprises no longer have the luxury of gradual preparation.
I don’t believe AI can replace systems that keep the world economy humming. But it certainly exposes whether they can operate at scale.
So what does it take to sustain scale right now?
This is the question I hear from every CIO I talk to. They know the problem. We’re all operating under a similar kind of sustained pressure. Their executive teams are telling them: you have to be AI-ready.
But you can’t boil the ocean. You have to pick a path, test it, and learn quickly, understanding where your systems hold and where they don’t.
Scaling without breaking will be critical. But just as important is turning that scale into growth. And those same forces pushing your systems to their limits can also create the biggest opportunities — if you are prepared to harness them.
We explore exactly what that looks like in our Scalability and Growth guide, from architectural patterns to operational guardrails designed for AI-driven demand.