Too many organizations bolt AI onto legacy systems and then wonder why nothing really changes. They graft new models onto aging platforms, plug chatbots into outdated workflows, and hope it counts as innovation. For more than 20 years, John Campbell Crighton has been writing a different playbook, one that starts with AI at the core and builds upward from there. As Chief Technology Officer at ONEngine, leading global engineering teams across healthcare, finance, and energy, Crighton has seen the same truth play out repeatedly: real leverage in AI comes from decisions made long before the models are trained. When the groundwork is rushed, or ignored entirely, AI turns into an endless series of costly fixes instead of a strategic advantage.
AI as a Backbone, Not an Add-on
The difference between AI as an add-on and AI as a backbone shows up immediately in how systems behave under pressure. When AI is tacked on at the end of a project, it’s disconnected and fragile. When it’s designed in from the beginning, it changes everything. At Caranta Solutions, Crighton’s team built a new EMR and revenue cycle management platform with AI embedded directly into the data layer from day one. AI supported clinician workflows, surfacing insights in real time and cutting administrative burden instead of adding more screens and clicks. “When AI comes at the end of the project, it becomes a patchwork. When it comes at the beginning, it becomes the backbone,” Crighton explains.
That architectural shift gave clinicians hours back every week by redesigning how the system processed, prioritized, and presented information.
This approach requires rethinking the entire stack. Data quality can’t be an afterthought when every AI decision depends on it. Infrastructure has to support computational demands without introducing new bottlenecks. And workflows must be designed so AI recommendations fit naturally into how people already work, rather than disrupting operations.
Infrastructure Reimagined
AI systems only crack when the underlying infrastructure can’t adapt. Traditional operating models struggle under AI workloads: batch processing slows response times, manual deployments become impossible as models evolve daily, and fixed infrastructure can’t keep pace with unpredictable demand. Without adaptability at the core, AI simply can’t deliver on its promise.“AI systems break if they’re not designed to breathe. They demand elastic infrastructure, automated deployment, and predictable release pipelines,” Crighton notes.
During large-scale AWS migrations, his teams moved to Kubernetes orchestration and CI/CD models, enabling reliable monthly updates. The shift wasn’t about chasing new tools; it was about designing systems that could grow without sacrificing speed, stability, or operational efficiency. The answer, he argues, is almost never “more resources.” It’s smarter architecture.
Container orchestration scales dynamically based on actual demand. Automated pipelines catch integration issues before they reach production. Monitoring tracks model performance alongside infrastructure metrics, surfacing degradation before users feel the impact. The result is an AI stack that can flex and evolve as quickly as the business does.
Teams That Build With Purpose
Even the best architecture fails if the people building it don’t understand why it matters. AI-driven systems demand collaboration across data, engineering, product, security, and compliance. Crighton has led teams of more than 100 people across time zones, industries, and regulatory environments. The pattern he sees is consistent: “The breakthroughs never came from hierarchy. They came from shared ownership,” he explains. When teams understand the real problems they’re solving for customers, they experiment more boldly, and ship more responsibly. Traditional organizational structures tend to fragment collaboration. By the time decisions move up and down multiple layers of hierarchy, opportunities disappear. Shared ownership shortens feedback loops, strengthens accountability, and aligns incentives around outcomes, not tasks.
The Difference Between Advantage and Expense
Over two decades of architectural decisions, Crighton has distilled three fundamentals that separate AI as a competitive advantage from AI as a cost center:
- Make AI foundational, not decorative. Incorporate AI into your architecture from the very beginning, not as a last-minute enhancement.
- Design for adaptability. Build systems—data, infrastructure, and workflows—that can support continuous refinement, new models, and changing business needs.
- Invest in strong, cross-functional teams. Even the most elegant design fails without teams capable of building, maintaining, and improving it over time.
“Nail those, and AI stops being a promise and becomes an advantage,” Crighton concludes.
For him, AI-first isn’t about chasing the latest technology trend. It’s about committing to an architectural redesign that makes intelligence foundational to how the organization operates.
The companies that will lead their markets in the next five years won’t be the ones with the longest list of AI tools. They’ll be the ones that made AI impossible to remove without dismantling the entire operation.
Want more of this thinking? Connect with John Campbell Crighton on LinkedIn for deeper insights on building AI-first software architectures that actually scale.
