The AI ambition gap: Why legacy systems can’t keep up

There’s a lot of energy pouring into AI right now. Everyone is running a pilot, experimenting, investing. But when I talk to leaders really trying to put AI into practice, the same pattern keeps coming up: their teams might be ready, but their infrastructure isn’t.

No matter the industry, systems are struggling to support the weight of AI. Data is locked in silos. Applications don’t talk to each other. Compliance makes change harder than it needs to be. And underneath it all, mainframes still run the world.

That experience aligns with what we’re seeing. In our recent survey of 550 product and technology leaders, 95% reported actively modernizing legacy systems, driven largely by the need to support AI and analytics at scale.

Gartner expects more than 40% of agentic AI projects to be canceled by 2027 because they’re too costly and can’t prove their value. The models tend to work. The surrounding systems and the way teams have to navigate them are what hold the work back.

Foundational modernization may not grab headlines, but it’s the difference between AI that scales and AI that stalls.

Why AI fails in legacy environments

Pilots work because they run in controlled conditions. Production is where the seams start to show. If your data lives in five systems that don’t agree on definitions, your model will drift. If your infrastructure can’t handle real-time load, latency will break the experience. If your models depend on fresh or fast-moving data, legacy pipelines simply won’t deliver it. And if security and compliance come in at the end, the release gets stuck in review.

It’s easy to prove the model. Putting it into production takes longer when the stack isn’t ready. That’s not an AI problem. It’s an engineering and operations problem that modernization solves.

There’s also a human cost when AI efforts stall. Engineers lose confidence in the roadmap. Product teams stop believing the organization can deliver what it promises. Early excitement turns into hesitation, and it becomes harder to get people energized about the next initiative.

When AI fails because the underlying systems can’t support it, teams start to question the strategy instead of the infrastructure. It creates cultural drag at the moment when organizations need momentum the most. Our research reflects this disconnect: 75% of product leaders say following through on product strategy remains a major barrier to success, especially in larger, more complex organizations.

Modernizing with purpose

Modernization doesn’t have to mean tearing everything down and starting over. The companies that move fastest take a more practical path. They start with the areas where change creates the most value. They focus on connecting what they already have instead of replacing it all at once. And they keep momentum by delivering improvements they can measure.

That’s the approach we take at Modus Create. We work inside existing environments and make progress where it has the most impact. Sometimes the right move is replatforming a core customer experience. Sometimes it’s automating a workflow that slows down every release, or bridging two legacy systems so teams can finally put their data to work. The pattern doesn't change: reduce friction, increase reliability, and make sure the underlying systems can support modern workloads, AI included.

Whatever the starting point, the goal is the same: intentional progress that strengthens the foundation without disrupting the business. When teams see their work move cleanly into production, confidence builds fast. Modernization gives them that traction and builds trust in the roadmap.

Engineering for scale, security, and compliance

In highly regulated industries, the hesitation around modernization isn’t about whether it should be done. It’s about making sure critical systems stay stable while the work happens. When you’re dealing with protected data, strict audit requirements, and complex approval paths, you can’t afford to move fast without being deliberate.

The cleanest path is to design for security and compliance from the start. That means building architectures that are secured by default, not after the fact. It means investing in data infrastructure that scales, logs, and audits reliably. And it means following processes that align with your industry’s requirements so teams aren’t stuck building workarounds later.

AI adds another layer of responsibility: protecting training data, tracking how models behave in production, and making sure outputs are traceable and explainable. It’s not theoretical—76% of organizations report that regulatory or ethical considerations have already slowed their AI deployments, often because governance wasn’t designed into the system from the start.

I’ve seen this approach work in some of the most regulated environments: healthcare, pharma, and financial services. One clinical research firm cut infrastructure incidents in half and made compliance reporting 40% faster by modernizing its cloud environment.

When modernization is designed with security and compliance built in, teams move faster without increasing risk. The end goal is an environment that can support AI in production, handle real-time workloads, and maintain the level of oversight their industry requires.

The real work of the AI era

AI itself isn’t falling short. It’s being held back by the systems and structures around it. The organizations that make progress are the ones willing to modernize their foundation first, even when that work is slow, complex, or invisible from the outside.

This modernization is both a technical shift and a cultural one. It asks teams to rethink how they build, secure, and operate their systems so AI can move from a pilot to something that delivers real value. And it requires leadership willing to support not just the systems, but the people navigating the change.

Modernization means letting go of familiar tools, changing long-standing habits, and building trust in new technologies. That emotional lift is often heavier than the technical one. To make it work, leaders need to invest in onboarding and retraining. They need to celebrate early wins to build momentum. And they need to stay close to the teams doing the work, so the process feels inclusive, not imposed.

In every industry I’ve worked in, the teams that invest in both their architecture and their people are the ones that actually get AI into production. The ones that wait often stay stuck in pilot mode, watching others pull ahead.

If AI is part of your roadmap, the best place to start might be the part no one sees—the foundation that makes everything else possible.

Learn how Modus Create’s modernization services can help your business →