The Case for Boring AI

Scott Davis has been inside more data environments than most people will ever see: Fortune 10 boardrooms, airline operations centers, manufacturing plants, and logistics hubs. He runs data and AI strategy for Lumenalta, holds MVP status with Databricks, and has spent much of the past decade at the intersection where enterprise ambition meets technological reality.

Scott is not trying to sell you on AI. He is often the person called in when the person who sold it to you was wrong. That is why his central thesis lands differently than much of what is being said about artificial intelligence today.

“If it’s not boring,” Davis says, “you’re doing it wrong.”

Your People Already Know What to Do

For all the conversation about AI strategy, Scott sees a surprisingly different problem inside most organizations. They are not short on ideas.

“Almost all the organizations I work with—and airlines are probably one of the most passionate ones—the people out in the operations know everything. They have a long list of things they want to do.”

The issue is not imagination or ambition. “The major problem is that the company has not enabled them to go do it.”

That distinction matters because the people closest to the work already see the inefficiencies. They know the workarounds, the processes that create friction, where information disappears, what customers complain about, and which systems no longer serve the business. AI has simply made it possible to imagine acting on that knowledge faster, and that is where the risk begins.

“The exciting nature of AI is one of the biggest risks,” Scott says.

When the organization does not provide the tools, infrastructure or permission to move, capable people find their own way forward. A business unit downloads a tool. Someone connects a database. A team builds an application. An experiment starts delivering something useful. It feels like progress, and it may actually be progress, but without the controls, security, production standards, quality measures, and governance necessary to support it, the organization has created something it cannot reliably scale, measure or trust.

That is rogue AI, and in Scott’s experience it is increasingly the consequence of an organization failing to enable people who are already trying to improve the business.

“Almost every single company I’ve ever worked with is struggling to figure out what to do foundationally,” he says. “To make achievable ambitions, you really have to deploy a strong, secure, governed foundation first.”

The Model Is the Wrong Conversation

Much of the public conversation around AI begins with the model: which one, how powerful, how fast, what it can do that the last one could not. Scott believes this is largely the wrong conversation.

“The real question is: how do I produce value at scale? And that never, ever comes down to a model.”

Models change. They can be replaced, upgraded, or swapped as the technology evolves. The enterprise capability surrounding them is harder to replace and far more consequential. Can an idea move into production safely? Can the system measure whether it is working? Can the organization understand what it costs, what value it creates, whether another team can use it, and whether someone can maintain it three years from now?

That is where AI strategy becomes operational reality.

Scott offers a deceptively simple metric for knowing when an organization is getting there: “If it’s not boring, you’re doing it wrong.”

A production-ready system, in his view, should begin to feel almost templated, built around parameters, quality checks, governance, and a reliable path to production. The goal is not a heroic technology project. It is a repeatable capability that becomes easier to use because the difficult work has already been done underneath it.

And the variable that ultimately tells you whether the system works?

“If you’re using it.”

Boring Is an Engineering Principle

Scott trained as a civil engineer before moving into data, and the discipline still shapes the way he thinks. Civil engineering is built around function.

“You build a water tower. There are not a million types of water towers. There are two. You can build it with a support structure or you can set it on the ground up on a hill. It just has to function.”

Technology, in his view, should eventually arrive at the same place: reliable, repeatable and useful. This is why some of the infrastructure developments Scott considers most consequential are also among the least glamorous.

He points to Delta Sharing, Databricks’ open data-sharing technology, as an example. Before tools like it, enterprise data sharing could mean FTP drops, custom APIs, portals, and dashboards, each accompanied by its own layers of access, security, maintenance and coordination.

Anyone who has worked inside a large organization knows the pattern. There is a dashboard, but it does not contain what you need. You request something different and someone promises another dashboard. You ask for access to the underlying data and the answer is no. The result is friction disguised as infrastructure.

Delta Sharing changed that logic by making governed access to data dramatically simpler. Scott’s point is less about the product itself than what it represents.

“It’s functional and you do it in 30 seconds and it works. There is no flashiness to that. It’s just really, really useful.”

This is what he calls gravity. The best infrastructure decisions eventually pull toward the same principles: less friction, greater efficiency, stronger governance, and easier access. The more mature the technology becomes, the less remarkable using it should feel.

How AI Failure Actually Happens

AI projects do not always fail dramatically. Often, they simply deteriorate.

A motivated team solves a problem quickly. A SQL query gets connected to a database, becomes an application, and gets pushed to an API. The problem has been solved, except the team did not build the controls around it.

A month later, the solution works a little less well. It is not exactly broken, but no one can say how much performance has changed. It may not be monitored. The organization may not know what it costs to operate or what value it is producing. What began as something useful is now something useful enough to keep running, but not structured well enough to understand.

Without the foundation, there are only two choices. Someone finds the time to go back and evaluate the system properly, or the system continues running. Most organizations know which one happens.

Scale that behavior across dozens or hundreds of business problems and the consequences compound.

“You have to manage every problem you’ve ever solved, continuously, across decades,” Scott says.

Eventually, excitement becomes maintenance and innovation becomes overhead. The application an engineer was excited to build becomes an IT support ticket someone else inherits years later.

“No engineer that’s developing something exciting ever wants to go build the boring parts,” Scott says. “But if you don’t do it, that super exciting thing that you did will eventually roll to an IT support ticket.”

That is how AI becomes technical debt, not because the original idea was bad, but because the boring parts were never built.

Aviation Has a Different Problem

For airlines, the challenge becomes even more complicated. Aviation is structurally a low-margin business. Investment cycles are uneven. Infrastructure is old because airlines themselves are old. And the operational consequences of technology failure can be significantly higher than in many other industries.

That creates an entirely rational form of conservatism.

“If something works and it’s old and you have no money, what do you do?” Scott asks. “Nothing. That’s exactly what you do.”

The problem is that AI does not wait for infrastructure modernization. The demand exists anyway. Employees see what is possible, teams experiment, business units find tools, and rogue systems begin filling the space between what people want to do and what the enterprise has enabled them to do.

Scott’s assessment of where airlines stand is intentionally provocative.

“I haven’t seen any airline really doing AI right yet.”

His point is not that airlines lack innovation. It is that even some of the industry’s most advanced organizations are still modernizing the data environments AI depends upon. And as we talked through that tension—an industry built on mature, dependable systems confronting a technology being sold almost entirely through the language of disruption— I found myself wondering whether we were actually talking about two very different versions of boring: boring old and boring new.

Boring Old vs. Boring New

The distinction stopped him.

“You just said something that I don’t think I ever rationalized in my head, even though I’ve said it out loud,” Scott said. “Boring and old is the current state; the new gets proposed as unknown and wildly exciting. But what should be proposed is boring and new.”

And there it was.

Legacy technology is already boring. It is mature, familiar and understood, and it runs systems the organization depends upon every day. New technology, by contrast, is almost always introduced through the language of transformation: exciting, revolutionary, game-changing, unknown.

That may be precisely the wrong way to position modernization inside a mature organization.

The objective is not to take a traditional organization and turn it into a technology experiment. It is to make it a mature, modern organization: still governed, still reliable and still trusted, but capable of doing substantially more.

“That’s literally the challenge in front of me every day,” Scott said. “How do I get a very mature, traditional organization to be a very mature, modern organization? Boring old and boring new.”

It is a simple distinction, but an important one. The goal of modernization is not excitement. It is to replace the limitations of old, dependable systems with modern systems that become equally dependable.

Trust Is the Real Prerequisite

Ask most leaders about AI and the conversation quickly moves toward potential. What could we automate? What could we predict? How much efficiency could we unlock? What could the business look like five years from now?

Scott believes another question sits underneath all of them: do you trust that any of it will actually happen?

“There’s a huge difference between potential and trust in it actually happening.”

This is the part of AI transformation that grand strategy documents struggle to solve. An organization can announce a billion-dollar opportunity, commission the roadmap, and form the advisory board, but somewhere inside the company leaders are still asking themselves whether the people, the process and the technology will actually work—and whether they will know if it does.

Trust is built when those questions become answerable.

Scott’s prescription is straightforward: start small, build observability in from the beginning and show the work.

“When you’re done with this tiny, small, little piece, you can show exactly what was put into it and what was gotten out of it.”

That is where executive confidence begins. Not with potential, but with proof. One small system works. The organization can see why it works. The cost is understood. The value is visible. The next use case becomes easier. Eventually, a billion-dollar AI vision stops looking like an abstract promise and starts looking like a roadmap.

By this point, “boring” had begun to mean something larger than reliable infrastructure. It was becoming a proxy for trust: the point at which technology has been tested, governed, and understood well enough that people can stop wondering whether it will work and start using it to do the work they already know needs to be done.

The Goal Is to Stop Thinking About It

Scott uses the automobile as an analogy. The earliest cars were exciting precisely because they were unreliable. Driving could become an adventure. The machine might break. You needed to understand how it worked or know someone who did.

Today, almost none of us think about the underlying technology. We get in, put the car in drive, press the accelerator and go. Nobody celebrates because the car started. That is maturity.

Scott recently spoke with a CTO whose organization had completed significant modernization work. When asked how things were going, the response was almost casual: “Oh yeah, it’s great. I don’t really think about it anymore. It just all works.”

For Scott, that may be the highest compliment a technology organization can receive. The CTO is calm, the infrastructure works, and nobody is talking about how exciting it is.

The Point Was Never the Technology

Before we ended the conversation, we asked Scott for three things he knows about almost every organization before he walks through the door.

The first is that nobody has time. People will make time to discuss what needs to change; finding the time to actually change it is another matter.

The second is that everybody is much less advanced than you think they are. Scott spends his career at the edge of data and AI. Most organizations do not. They are running airlines, factories, logistics networks and businesses, and the knowledge gap is almost always larger than outsiders assume.

The third is the one that matters most.

“Every single organization has people that are absolutely passionate about what they do. And the number one thing they want is to be enabled to go do it.”

This is particularly visible in aviation. People spend entire careers in the industry. Twenty-year tenures are unremarkable. Institutional knowledge sits everywhere: in operations centers, commercial teams, maintenance organizations, airports, network groups, and countless other corners of the business.

These people already know where the opportunities are. They do not need another presentation explaining that AI will change the world. They need the ability to act on what they already know.

“AI is the single most impactful enabler that has ever existed,” Scott says. “If you could just give people the ability to go do what they already know how to do and put it into production, you would see a massive change—in efficiency, in satisfaction, in the overall experience of working in the aviation space.”

That is the larger argument behind boring AI. The foundation is not the destination. Infrastructure is not the point. Governance is not the point. Even AI is not really the point.

The point is creating an organization in which capable people can move from knowing what should be done to actually doing it safely, repeatedly, and at scale. For all the extraordinary promises surrounding artificial intelligence, getting there may require something far less exciting.

“Until all of the foundational steps are in place—mature, boring, trusted, secure, governed—you can’t enable anybody to go do anything.”

That is the blocker. And that is the work.